 

q
. L .. . ,
V&Lm$5=.:.: .
.4VVMMWW V. 3W2 \nﬁ ..

 

«
113m V
.c. .

mu ,

 

I..- .13V.%hq

«an. 1 . . u.
2 h. dawn“...
slznﬁa

.fi 1...
.173... A
.a!!:...v!.

. [Alb
O. ..§1I .

.xula

"by .égmmngh

vll‘... ﬁn:

 

3-5.92.5: by a
, 2:

._...y.._.,.:uma.au 3» n

 

“+5818

(m (‘0

 

 

 

 

 

LIB RARY lllllllllllllllllllllllllllllllllllll
Michigan State 3 1293 01766 2986
University

 

 

 

This is to certify that the
dissertation entitled
DEREGULATION IN THE POST-DIVESTITURE
LONG-DISTANCE TELECOMMUNICATIONS MARKETS

presented by

David Eugene Burnstein

has been accepted towards fulﬁllment
of the requirements for

PhD. degree in Economics

Wallace “l7- Ma

Major professor

 

Date avg-454: [0: \‘i‘l8

MSU is an Afﬁrmative Action/Equal Opportunity Institution 0-12771

 

 

 

 

PLACE IN, RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1“ WM“

DEREGULATION IN THE POST-DIVESTITURE LONG-DISTANCE
TELECOMMUNICATIONS MARKETS

By

David Eugene Bumstein

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Economics

1998

ABSTRACT

DEREGULATION IN THE POST-DIVESTITURE LONG-DISTANCE
TELECOMMUNICATIONS MARKETS

By

David Eugene Bumstein

Following the 1984 divestiture, state and federal regulators implemented a variety
of (de)regulatory policies to facilitate the transition to a market environment. In the ﬁrst
two essays, presented in chapters two and three, I discuss the eﬂects of these policies on
the inter- and intrastate long-distance markets’ performance, respectively. The ﬁnal
essay, presented in chapter four, examines the causes of rate deregulation in the intrastate
markets.

The principal results of the interstate study are the following: First, the steep
decline in access fees and development of “equal access” technology were responsible for
a signiﬁcant portion of the decline in interstate prices. The equal access policy was also
responsible for a substantial decline in the carriers’ market power. Second, the Federal
Communications Commission’s implementation of price-cap regulation had an economic
and statistically insigniﬁcant impact on carriers’ prices and market power. Third, the
subsequent rate deregulation of AT&T’s domestic interstate services contributed to the
recent increase in the carriers’ prices and restored some of their market power. Finally,
the estimates of AT&T and its competitors’ Lerner indices of market power were 0.64
and 0.34, respectively. These results suggest the carriers still possess signiﬁcant market

power.

 

First. i
5 pert

based

comp

incre.

char.
Base

follo

dere
sigr

den

pr:-

\\ c

 

Tl:

Ex,

 

 

The intrastate study reached similar conclusions as the preceding interstate study.
First, implementation of rate deregulation reduced carrier “competitiveness” producing a
5 percent increase in the market price. Second, an increase in the number of facility-
based long-distance carriers competing in a state had the perverse effect of reducing
carrier competitiveness and generating an 8 percent increase in the market price. Third,
in those states that deregulated entry to the adjacent intraLA TA toll market, the
competitiveness of interLA TA market conduct diminished, resulting in a 5 percent
increase in the interLA TA market price.

The general objective of the ﬁnal essay is to ascertain political and economic
characteristics of a state that were conducive to the implementation of rate deregulation.
Based on a private interest theory of economic regulation the study uncovered the
following: First, the intensity of business interests, measured as the percentage of
business to total local access lines, was positively correlated with the probability of rate
deregulation. Second, the prior employment characteristics of the regulators was a
signiﬁcant factor in the selection of rate deregulation, that is, the probability of rate
deregulation decreased with the percentage of commissioners previously employed as
(residential) consumer advocates and increased with the percentage of commissioners
previously employed as senior level business managers or owners. Finally, two tests
were conducted for (i) the presence of regulator slack and (ii) a “demonstration effect.”
The former tested for the product of slack, or “shirking” activity, while the latter
examined the regulated ﬁrm price setting behavior for a strategic component. The results
of the tests were unable to reject shirking and strategic ﬁrm behavior as inﬂuential

factors.

To Stephanie, whose love, strong support, motivation, warmth, and ability to leap
small buildings in a single bound create a partnership with the freedom of spirit to

explore, grow, and achieve.

of

M

CO

I\

M

 

 

ACKNOWLEDGMENTS

I extend sincere thanks to a number of signiﬁcant contributors to the completion
of this work. First, I owe an enormous debt to my committee members, Drs. Wallace
Mullin, Kenneth Boyer, and Harold Trebing. Their support and guidance were
invaluable. My respect and gratitude ﬂow easily when expressing appreciation to the
committee’s chairman, Dr. Mullin. His exceptional knowledge and wise counsel are gifts
I will always remember and value.

Last and most important, I acknowledge and offer heartfelt thanks to Eugene and
Martha Kathryn Bumstein. They have the rare ability as parents to provide unconditional
love and an unwaning expectation that you will accomplish as professionals and as

people. In many moments and in many ways, they have been my inspiration.

LIE

CH

CH

CH

 

TABLE OF CONTENTS

LIST OF TABLES ........................................................................................................... viii
LIST OF FIGURES ......................................................................................................... x

CHAPTER 1. THE POST-DIVESTITURE LONG-DISTANCE MARKET
ENVIRONMENT: INTRODUCTION AND OVERVIEW ..................... 1

CHAPTER 2. REGULATORY CHANGE AND ITS IMPACT ON LONG—
DISTANCE PRICES AND MARKET POWER: THE INTERSTATE

MARKET .................................................................................................. 9
2.1 INTRODUCTION ............................................................................................... 9
2.2 BACKGROUND ................................................................................................ 14
2.3 A MODEL OF FIRM DEMAND AND CONDUCT ......................................... 21
2.4 THE EMPIRICAL RESULTS ........................................................................... 30
2.5 CONCLUSION .................................................................................................. 41
APPENDIX 2A .......................................................................................................... 43
APPENDIX 28 .......................................................................................................... 45
APPENDIX 2C .......................................................................................................... 48
APPENDIX 2D .......................................................................................................... 50
APPENDIX 2E .......................................................................................................... 51

CHAPTER 3. REGULATORY CHANGE AND ITS IMPACT ON LONG—
DISTANCE PRICES AND MARKET POWER: THE INTRASTATE

MARKET ................................................................................................. 52

3.1 INTRODUCTION .............................................................................................. 52
3.2 STATE MOTIVES FOR IMPLEMENTING ALTERNATIVE

REGULATION .................................................................................................. 58

3.3 LITERATURE REVIEW ................................................................................... 60

3.3 A MODEL OF MARKET DEMAND AND CONDUCT ................................. 63

The General Model .................................................................................. 63

The Regulatory Experiment Model ......................................................... 72

3.4 THE EMPIRICAL RESULTS ........................................................................... 78

3.5 CONCLUSION .................................................................................................. 87

APPENDIX 3A .......................................................................................................... 90

APPENDIX 38 .......................................................................................................... 93

vi

CHAPTER 4. AN EMPIRICAL TEST FOR THE PRESENCE OF POLITICAL
SLACK AND STRATEGIC FIRM BEHAVIOR IN THE TRANSITION
TO RATE DEREGULATION IN THE INTRASTATE LONG-

DISTANCE TELECOMMUNICATIONS MARKETS .......................... 94
4.1 INTRODUCTION ............................................................................................... 94
4.2 LITERATURE REVIEW ................................................................................. 100
The Theory of Economic Regulation ..................................................... 100

Empirical Analysis of Regulatory Change: the Telecommunications
Industry ............................................................................................... 102
4.3 A MODEL OF REGULATORY TRANSITION ............................................. 104
Description of the Model ....................................................... 104
Testable Implications of the Model ....................................................... 110
4.4 Empirical Analysis ............................................................................................ 112
Description of the Variables: Logit equation (4.6) ................................ 116
Description of the Variables: OLS equation (4.7) ................................. 120
4.5 EMPIRICAL RESULTS ................................................................................... 121
4.6 CONCLUSION ................................................................................................. 128
APPENDIX 4A ........................................................................................................ 130
APPENDIX 48 ........................................................................................................ 133
APPENDIX 4C ........................................................................................................ 134
APPENDIX 4D ........................................................................................................ 138
APPENDIX 4E ........................................................................................................ 139
REFERENCES .............................................................................................................. 140

vii

TA

TA

TA

TA

TA

TA

TA

TA

 

TABLE 1:

TABLE 2:

TABLE 3:

TABLE 4:

TABLE 5:

TABLE 6:

TABLE 7:

TABLE 8:

TABLE 9:

TABLE 10:

TABLE 11:

TABLE 12:

TABLE 13:

TABLE 14:

LIST OF TABLES

ESTIMATES OF MARKET POWER IN THE INTERSTATE

LONG-DISTANCE TELECOMMUNICATIONS MARKET ................ 10
VARIABLE NAMES, DEFINITIONS, AND DESCRIPTIVE
STATISTICS ........................................................................................... 26
TWO-STAGE LEAST SQUARES ESTIMATES: INVERSE

DEMAND FUNCTIONS ......................................................................... 31
TWO-STAGE LEAST SQUARES ESTIMATES: SUPPLY
RELATIONS ........................................................................................... 34
EMPIRICAL ESTIMATES OF MARKET POWER .............................. 39
APPENDIX 2A: ESTIMATES OF a ..................................................... 44
APPENDIX 23: NONLINEAR TWO-STAGE LEAST SQUARES
ESTIMATES — SUPPLY RELATIONS ................................................. 46
APPENDIX 2C: FIRST-STAGE ESTIMATES — DEPENDENT
VARIABLE InQi ...................................................................................... 48
APPENDIX 2C: FIRST-STAGE ESTIMATES - DEPENDENT
VARIABLE Q,/Qj .................................................................................... 49
APPENDIX 2D: ORDINARY LEAST SQUARES ESTIMATES ........ 50
APPENDIX 2E: POST-DIVESTITURE TIMELINE ............................ 51

RECENT CHANGES IN THE REGULATORY POLICY OF 36
MULTI-LATA STATES ......................................................................... 53

A MATRIX DESCRIPTION OF THE REGULATORY EXPERIMENT
MODEL ................................................................................................... 56

VARIABLE NAMES, DEFINITIONS, AND DESCRIPTIVE
STATISTICS ........................................................................................... 7O

viii

TAB

TAB

TAB

TAB

TAE

TAT

TAI

TAl

TA

 

TA]

TA

TA

TA

 

 

TABLE 15:

TABLE 16:

TABLE 17:

TABLE 18:

TABLE 19:

TABLE 20:

TABLE 21:

TABLE 22:

TABLE 23:

TABLE 24:

TABLE 25:

TABLE 26:

TABLE 27:

TABLE 28:

TABLE 29:

TWO-STAGE LEAST SQUARES ESTIMATES — INVERSE DEMAND

FUNCTIONS ........................................................................................... 80
TWO-STAGE LEAST SQUARES ESTIMATES: SUPPLY

RELATIONS ........................................................................................... 82
SUMMARY OF RESULTS .................................................................... 84

APPENDIX 3B: NON-LINEAR LEAST SQUARES ESTIMATES -—
SUPPLY RELATIONS ........................................................................... 93

VARIABLE NAMES, DEFINITIONS, AND DESCRIPTIVE

STATISTICS ......................................................................................... 1 14
Two-STAGE ESTIMATES: OLS EQUATION ................................. 123
TWO-STAGE ESTIMATES: LOGIT EQUATION ............................. 125
APPENDIX 4A: VARIABLE DESCRIPTIVE STATISTICS ............. 130
APPENDIX 4A: TWO-STAGE ESTIMATES — OLS EQUATION 131
APPENDIX 4A: TWO-STAGE ESTIMATES — LOGIT

EQUATION ........................................................................................... 132
APPENDIX 48: FIRST-STAGE ESTIMATES ................................... 133
APPENDIX 4C: TWO-STAGE ESTIMATES — TOBIT

EQUATION ........................................................................................... 136
APPENDIX 4C: TWO-STAGE ESTIMATES — OLS EQUATION... 137
APPENDIX 4D: TWO-STAGE ESTIMATES - STATE DUMMY
VARIABLES ......................................................................................... 138

APPENDIX 4E: TWO-STAGE ESTIMATES - UNRESTRICTED
LOGIT EQUATION .............................................................................. 139

FlGL’Rl

FIGL'Rl

FIGL’RI

LIST OF FIGURES

FIGURE 1: REAL CONSUMER PRICE INDEX FOR INTERSTATE
LONG-DISTANCE SERVICE ................................................................. 9

FIGURE 2: AVERAGE INTERSTATE, REAL ACCESS CHARGE FOR A
MINUTE TELEPHONE CALL .............................................................. 16

FIGURE 3: THE DEVELOPMENT OF EQUAL ACCESS IN THE INTERSTATE
LONG-DISTANCE MARKET ............................................................... 18

by the E
divestitt
operatic
to dives
Compar
services
previous
created ;
35 Well 2
The rest
PTOVIdin

Chapter One

The Post-Divestiture Long-Distance Market Environment:
Introduction and Overview

The 1982 Modiﬁed Final Judgment (MF J) - which resolved an antitrust suit ﬁled
by the Department of Justice in 1974 and formed the basis for the January 1, 1984
divestiture of AT&T - allowed AT&T to keep its long-distance operations, manufacturing
operation, and research and development facility. However, the judgment required AT&T
to divest its monopoly local exchange business by spinning off its Bell Operating
Companies. In return, AT&T was permitted to enter businesses beyond common carrier
services, such as the computer industry, which it had been prohibited from doing
previously. The seven independent Regional Bell Operating Companies (RBOCS),
created from the Judgment, were to provide local exchange services to the end-customer
as well as provide exchange access to the long-distance interexchange carriers (IXCS).
The restrictions placed on the RBOCS were as follows: (1') they were forbidden from
providing interlata long-distance service; (ii) they could not manufacture equipment; and
(iii) they could not provide information services.l

In addition to this dramatic industry-wide restructuring, ﬁve developments
speciﬁc to the long-distance markets transpired concurrent or subsequent to divestiture:

(i) the implementation of “equal access” technology, (it) non-regulated entry to the long

 

’ LATA, short for Local Access and Transport Area, is a geographic entity created at the time of divestiture
as the demarcation boundary for the RBOCS. Calls which cross the LATA boundary, interLATA calls,
must be passed off to a long distance company. IntraLATA calls do not cross a LATA boundary and, thus,
are typically transmitted end-to-end by the RBOC.

 

 

 

distance r
expansior

Fi
distance c
AT&T to'
IXC. Thc
among th:

Sc
eliminatir
the mark:
Commiss
every app

T1
prices pril
access Ch;
The OUICC
ofan 3CCc
Pending u—
the local t
rebajan Cir

Pﬁs On a:

K
see KaStﬁ'

 

 

distance markets, (iii) rate restructuring, (iv) incremental rate deregulation, and (v) the
expansion of ﬁber optic transmission technology.

First, a provision in the Judgment required the RBOCS to provide all long-
distance carriers with “equal access.” This provision enabled a customer to make a non-
AT&T toll call without dialing several extraneous numbers to connect to his preferred
IXC. Therefore, once an RBOC had implemented equal access, a customer could choose
among the available IXCS by simply presubscribing to his preferred carrier.

Second, during the period of divestiture both state and federal regulators were
eliminating virtually all regulatory barriers into long-distance markets. That is, entry into
the market only required ﬁling an application with the appropriate state Public Service
Commission (PSC) and the Federal Communications Commission (FCC), with “virtually
every application for entry being approved.”2

Third, in compliance with the MF J , the FCC began rebalancing local and toll
prices primarily through two related activities. The ﬁrst pertains to the FCC’s 1984
access charge ruling which mandated aifundamental change in RBOC access charges.
The outcome was a substantial decline in the access fees paid by IXCS and the application

3
” The second

of an access fee to the ﬁnal consumer via a “subscriber line charge.
pertains to a series of accounting changes instituted by the FCC which shifted a portion of
the local exchanges’ ﬁxed costs from interstate to intrastate telephone service. The

rebalancing of rates was concluded in 1992, when the FCC stopped requiring AT&T to

pass on access fee reductions to its customers.

 

2 See Kaserman and Mayo (1994), p.89.

 

SBI'VICI
intra- :
the 38
maon
regulaI
states t
deregu
regulat
rate-Of-
regular.
be reclz

the derc

dramati.
it Was b,
micro“;
and requ

Conciudc

See M73
251.231
‘ 'ﬁ.
See Kfide
{CEUlaIOr}
e Mom,

PNCe.Cap re

 

Fourth, while incremental changes in the rate regulation of intrastate long-distance
service followed divestiture, wide-spread rate deregulation has transpired in both the
intra- and interstate markets in more recent years. For example, between 1984-89, 28 of
the 38 multi-LATA states eliminated traditional rate-of-return regulation, with the
majority of these states choosing a pricing-ﬂexibility regulatory policy (e.g., price-cap
regulation, banded rate-of-return regulation, moratorium regulation, etc.)4 and just a few
states choosing rate deregulation. In contrast, subsequent to 1989 the pool of rate
deregulated states has expanded dramatically, while the number of pricing-ﬂexibility
regulated states has declined. AS of present all states have ceased imposing traditional
rate-of-return regulation. In the interstate market, the FCC replaced rate-of-return
regulation with price-cap regulation in July 1989. The FCC granted AT&T’S request to
be reclassiﬁed as a “non-dominant” common carrier in October 1995. The outcome was
the deregulation of domestic rates for interstate long-distance service.5

Finally, the rapid development and expansion of ﬁber optic technology
dramatically altered the method of long—distance transmission. At the time of divestiture
it was believed that long-distance carriers would compete against each other using
microwave technology. Since microwave transmission does not entail large sunk costs
and requires minimal trafﬁc before it reaches diminishing returns to scale, it was

concluded that the long-distance market was not a natural monopoly. Hence, it was

 

3 See M725 and WA TS Market-Structure, Third Report and Order, CC Dkt. No. 78-72, 93 F.C.C. 2d 241,
251,252.

’ See Kridel, Sappington, and Weisman (1996), pp. 271-274, for a detailed description of pricing-ﬂexibility
regulatory policies.

5 See Motion of A T& T C orp. to be Reclassiﬁed as a Non-Dominant Carrier, 1 1 FCC Red 3271, 3288, 1|26
(1995). The FCC still imposes rate averaging based on time of day, distance and duration. In addition,
price-cap regulation is still maintained in the international telecommunications market.

 

 

 

micro“:
market.

compan'
ﬁber—op
transmis
long-dis
network
cumbers
minimai
II was CI
large se;
Up as ne
SalVage

0f ﬁber

aCllUStm.
IO Linkn

dit'estiru

microwave radio technology that was an impetus to the deregulation of the long-distance
market. “But even as the ink was drying on the divestiture decree, long-distance
companies were dynamiting their microwave towers to the ground, replacing them with
ﬁber-optic glass.”6 The rise of ﬁber optic cable as the dominant technology for the
transmission of interexchange services dramatically changed the cost structure of the
long-distance market. The labor and capital required to construct a nationwide ﬁber
network was substantial and the acquisition of the necessary rights of way was legally
cumbersome. The incremental costs of installing cable with additional optical ﬁbers was
minimal, yet the incremental beneﬁt from increased capacity was potentially large. Thus,
it was considered efﬁcient to lay the groundwork for an optical ﬁber network, in which
large segments were initially dark, and as the market developed the segments would be lit
up as needed.7 Lastly, once a network was built it had few alternative uses (i.e., its
salvage value was close to zero).8 Therefore, the long-distance infrastructure composed
of ﬁber optic glass is structurally quite different than a microwave based network.

The implementation and expansion of equal access, non-regulated entry, rate
restructuring and deregulation, and ﬁber optic technology were important structural
adjustments that signiﬁcantly altered the long-distance market environment. According
to Lawrence Garﬁnkel, AT&T’S former vice president of public affairs, these post-

divestiture developments were “unprecedented events in business history: All of the

 

6 See Huber (1993a), p.A10.

7 “Dark” and “lit” are industry terms used to distinguish between portions of the ﬁber network that are
connected to functioning electronics (“lit”), and portions of the network not in use (“dark”).

3 See Huber (1993b), p.36. For an argument contesting the claim that ﬁber optic networks are a barrier to
entry see Kaserman and Mayo (1994), p.90.

busine
opinio
"estab‘.
remox'
pricing
compe
[Bell (
liuber
"If one

marke‘

comp-e

of the

AT&T

business in a mature market has suddenly been thrown up for grabs.”9

In his written
opinion, Judge Harold Greene stated that the fundamental objective of the MF J was to
“establish the foundation for a truly competitive telecommunications industry...with the
removal of these barriers to competition, AT&T should be unable to engage in monopoly
pricing in any market.”'0 Five years later, Judge Greene again concluded “that
competition now exists in the interexchange market, and that the entry of the Regional

1 1
” Peter

[Bell Operating] Companies into that market is not necessary to give it Vitality.
Huber, the author of the Department of Justice’s 1987 review of the MF J , conﬁrmed that
“If one is to judge by installed transmission capacity, competition in the interLATA
markets is healthy...[and] in many large urban markets apparently remains vigorous.”12

During much of the post-divestiture period, statements hailing the burgeoning
competition in the long-distance market were solidly supported by the statistical proﬁle
of the industry. For example, not only were AT&T’S long distance rates reduced about
45% in real terms between 1984 and 1991, but the company’s share of the long-distance
market dropped from 84.2% in 1984 to 62.9% in 1990 as well. Demand for switched
access minutes grew 111% from 37.5 billion in 1984 to 79.1 billion in 1991. While
AT&T’S volume increased by 58%, that of its competitors rose almost 400% during this
same time frame.‘3

Unfortunately, this apparent trend towards a competitive long-distance market has

recently shown Signs of regression. For example, in 1992, while access fees continued to

 

9 See Garﬁnkel (1993), p.325.

‘° United States v. AT&T, 48 PUR 4th 227, 552 F.Supp. at 172 (D.D.C. 1982).
“ 673 F. Supp. at 550.

'2 See Huber (1987), p.3.2.

'3 These statistics were taken from Noam (1992), p.443.

 

 

decline.
time sir
develop
evolx'in
regulati
particul
distance
Taylor 1
more th
distance
other cc
AT&T'
fees.IS

111610“.
“'35 ma;

and Set _

teIECQm

SUUCIUT;

14\
: See Hui

“Premn
Ong dish:
1‘ the pht
number to

See Ma‘

1

decline, the basic rate for residential interstate telephone service increased for the ﬁrst
time since divestiture and has continued to increase in each subsequent year. This
development has given support to those asserting that the long-distance market is not
evolving towards a competitive market, but is instead “a stable oligopoly, propped up by

14
as In

regulation, and operating under an AT&T-supplied canopy of umbrella pricing.
particular, it is argued that the decline in AT&T’S market Share and the price of long-
distance service are largely the result of regulatory mandate. For example, Taylor and
Taylor (1993) indicate that the overall reduction in interstate long-distance prices was
more than explained by the reduction in the carrier access charges paid by the long-
distance carriers to the local telephone companies. Likewise, in an attempt to make the
other common carrier’s (OCC’S) “non-premium” interLATA service competitive with
AT&T’s “premium” service, the FCC mandated a 45% reduction in non-premium access
fees.15 It is held that this disparity in access fees more than compensated the OCC’S for
the lower quality service they provided. In fact, the higher access charges that AT&T
was mandated to pay, in effect, raised its costs relative to those of the other large carriers
and set in place conditions that forced shifts away from AT&T in market share.16
Therefore, given the post-divestiture regulation of the long-distance

telecommunications market, it is not unusual to see dramatic changes in the market’s

structure. However, after taking into account the impact of these new policies, it is

 

" See Huber (1992), p.1.15.

'5 “Premium” and “non-premium” access are used to describe the type of access that a customer has to the
long distance network. Premium access, often referred to as “1+” service only requires the customer to dial
1+ the phone number of its desired destination. Non-premium access requires a customer to dial a separate
number to contact his long-distance provider, enter an identiﬁcation number and then dial the number of
the desired destination.

'6 See MacAvoy (1995), p.152.

argued that the long-distance market is not evolving towards a competitive market, but
instead is becoming a natural monopoly with signiﬁcant barriers to entry. Peter Huber,
who in 1987 was a proponent of “vigorous” long-distance competition, cogently restated
his position in a 1993 Regulation article:

Despite all the hype from Candice Bergen, the long-distance industry is now
characterized by umbrella pricing, under a canopy maintained by FCC regulation
and AT&T. AT&T is ostensibly subject to ‘price cap’ regulation by the FCC,
but the cap is really a ﬂoor. The Commission spends most of its time making
sure that AT&T does not lower its prices too fast; competitors rush to court
whenever AT&T’S prices seem likely to fall...Under careful FCC supervision,
AT&T pays off MCI and Sprint in market Share and subsidized local access, in
return for which they shield AT&T from the government. The payoff may seem
high - $8 billion a year to MCI alone - but it protects AT&T’s $53 billion in
assets and $63 billion in annual revenues from further predation by lawyers. No
insider is going to be crazy enough to disrupt this cozy little arrangement. What
we have here is umbrella-priced oligopoly, in which everyone is kept calm by the
combined wisdom of one market leader and an acquiescent federal
commission. 1 7

The chapters that follow attempt to contribute empirical evidence on the effects (in
chapters 2 and 3) and causes (in chapter 4) of post-divestiture deregulation in the intra-
and interstate long-distance markets. Each of the three chapters are self-contained essays
that stand on their own. Taken together, however, they paint an interesting picture on the
successes and failures of economic regulation. Before proceeding the objective of each
chapter is brieﬂy outlined below.

Chapters two and three examine the impact of regulatory change on the long-
distance carriers’ prices and market power. This is accomplished by implementing static
oligopoly models that nest various types of ﬁrm and market conduct. In both chapters the

data is disaggregated into time-period subsamples based on the occurrence of

 

'7 See Huber (1993b), p.38.

 

 

hypoth
\aﬁaﬁi
regulat
interst:

market

variabl
issues I
behavi.
policy

change
econon
suggeg
legit ec

ex21min

hypothesized changes in the market structure. More speciﬁcally, the data is tested for
variation in the ﬁrm and market conduct variables in response to changes in the
regulatory environment. Chapter two examines ﬁrm level time-series data on the
interstate market, while chapter three examines state level panel data on 36 intrastate
markets.

Finally, chapter four examines the inﬂuence of a variety of economic and political
variables on the probability that a state chooses to implement rate deregulation. Two
issues are given particular attention. First, I test for the presence of strategic ﬁrm
behavior in a pricing-ﬂexibility regulatory environment. This requires that the regulatory
policy and the regulated price be recognized as endogenous to the model of regulatory
change. A simultaneous (OLS-Logit) system based on a private interest theory of
economic regulation is speciﬁed. Second, I test for the presence of political Slack, as
suggested by Kalt and Zupan (1984). In doing so, I include a vector of variables in the
logit equation that control for the cost of (Slack induced) regulator “shirking” and

examine the model’s response to the inclusion of these variables.

 

 

2.1 lntror

W.
distance 8
consistent
asymmetr
period In:

explanati

See “! a;

 

Chapter Two

Regulatory Change and Its Impact on Long-Distance Prices
and Market Power: The Interstate Market

2.1 Introduction

While there has been a substantial overall decline in the prices of interstate long-
distance service since 1984, in recent years these prices have demonstrated a gradual but
consistent upward trend (see Figure 1). Unfortunately, the numerous, often complex and
asymmetric, changes in regulatory policy that have occurred during the post-divestiture
period make it difﬁcult to Specify a cause. A not surprising result is that the various

explanations Offered tend to conﬂict.l

Figure 1:
Real Consumer Price Index For Interstate Long-Distance Service

(CPI Indexz1982-84 I 100)

100

 

80

60

 

Index

40 7777 7 . 7 7 77

20 ___..__._.777. 77 7 ,7 .,7 -7 _ -7 . 7_77-_ .77

 

 

 

1984 1986 1988 1990 1992 1994 1996
YEAR

 

' See Ward (1995), MacAvoy (1995), Kahai, et. al. (1995), and Knittel (1997).

Study

Ward (1 995)
Interstate,
Regional Datab
Intrastate,
Regional Datab

Table 1:

Estimates of Market Power in the Interstate Long-Distance Telecommunications Market

AT&T’s Lerner Sprint and MCI’s
index Lerner index

0.16 0.07

0.19

 

Kahai, et.al. (1996)
Interstate,
National Data”

0.13/0.29'

 

MacAvoy (1996)
Interstate,
Regional Datab

 

Knittel (1997)
Interstate,
National Datab

 

Market power based on AT&T’S share of interstate output and industry assets,

respectively.

b Price data based on Message Toll Service (MTS), a switched long-distance service
used primarily by residential consumers.
c Lerner index based on the weighted average price of AT&T, MCI and Sprint.

 

The ﬁndings of Ward (1995) and Kahai, et. a1. (1996) are based on various
models of oligopoly market conduct. Examining the tariffed prices of the long-distance
carriers, the authors derive estimates of AT&T, Sprint, and MCI’S Lerner indices of
market power.2 Both papers conclude that long-distance service prices have converged to
their competitive levels; while there is some variation in their estimates, in general the
price markups are found not to exceed 20 percent, indicating a lack of signiﬁcant market
power (see Table 1). In achieving these results both papers impose signiﬁcant restrictions
on the market conduct of the long-distance carriers. For example, Ward’s model restricts

carrier behavior to be consistent with Coumot equilibrium conduct and Kahai, et. al.

2 Carrier i’s Lerner index is deﬁned as .4, = (P,- - MC,- )/P,- , where P; and MC; are ﬁrm i’s price and
marginal cost, respectively.

10

impose the Dominant F inn/Competitive Fringe (DF/CF) model, which assumes AT&T’S
rivals are price takers while AT&T is the sole price maker.3 In addition, both papers
restrict the assumed behavioral interaction to be unchanged over the sample period.

Based on premises that differ from Ward and Kahai, et. al., MacAvoy (1995) and
Knittel (1997) conclude that performance in the long-distance market diverges
signiﬁcantly from a competitive outcome; examining tariffed prices, the authors derive
Lerner indices in excess of 50 percent, indicating substantial market power (see Table 1).
The fundamental assumption made by MacAvoy and Knittel is that marginal costs are
measured with sufﬁcient accuracy from accounting data, which in turn are used directly
by the researcher to calculate market power. MacAvoy attributes the substantial price
markups to tacit collusion among the long-distance carriers. His empirical evidence
comes from regressing the weighted average Lerner index of AT&T, MCI, and Sprint on
the Herﬁndahl index of market concentration.’ MacAvoy discovers a negative
relationship between market power and concentration. Since we would expect to see the
opposite relationship in a non-cooperative oligopoly market, he concludes that tacit
collusion is what is driving market power.

Knittel’s hypothesis is that the substantial markup of long-distance prices is
derived from customer search and switching costs. He regresses a carrier’s Lerner index
on the variables FEE and STDRA TES, where FEE is the amount charged by local carriers

when a customer switches long-distance providers and STDRA TES is the standard

 

3 The DF/CF model also imposes product homogeneity (i.e., a single market price for interstate long-
distance service). This is an exception to the literature reviewed in this paper. Likewise, the model
presented in section III is generalized to allow for product heterogeneity.

11

deviation
present 11
the lone:
support I

deterrnin

impleme
perfect c
does not
Ward ar
the data
L'nfortu
such as
Chilnge.
accordi:
Speciﬁc

reglllatr

from M

deviation of rates for a ﬁrm in a given year. FEE is a proxy for the switching costs
present in the market and STDRA TES captures the costs of searching for a carrier Offering
the lowest price. The positive and highly Signiﬁcant coefﬁcients on these variables
support the author’s hypothesis that search and switching costs are signiﬁcant
determinants of market power in the long-distance market.

In this paper I examine market power in the long-distance market by
implementing a static oligopoly model that nests various types of market conduct from
perfect competition to Coumot and joint proﬁt maximization. In particular, the model
does not restrict market conduct to be consistent with a single theory of oligopoly, as do
Ward and Kahai, et. al., but derives an estimate of market conduct that is consistent with
the data. Further, Ward and Kahai, et. a1. assume market conduct is time invariant.
Unfortunately, this simplifying assumption is overly restrictive when examining markets,
such as long-distance telecommunications, that are experiencing signiﬁcant structural
change. For this reason I disaggregate the time-series data in this study into subsamples
according to the occurrence of hypothesized changes in market structure. More
speciﬁcally, I allow for market conduct to vary over time in response to changes in the
regulatory environment.

Finally, to assess the robustness of the model, and corroborate the results derived
from MacAvoy and Knittel, I examine the model subject to two speciﬁcations of
marginal cost: (1) complete cost information — each component of the marginal cost

function is observed by the econometrician; (2) incomplete cost information — one or

 

’ The Herﬁndahl index is deﬁned as H = 2:4 MS,-2 , where MS; is the market share of the 1‘h ﬁrm in a

12

 

 

more 1
specifi
ﬁrnttc
under
corntXJ

Chan

”1
n1

dunng
caniers
Sectnan

summaj

access”

 

more components of marginal cost are not observed by the econometrician. Both
speciﬁcations require the functional form of marginal cost be properly speciﬁed and the
ﬁrm to have complete information on this function. The analysis of market conduct
under speciﬁcation (2) is made tenable by using information on changes in the observed
components of marginal cost (such as a decrease in access fees), plus the observed
changes in the equilibrium outcomes.5

The following section describes the FCC’S (de)regulatory policies implemented
during the post-divestiture period and the hypothesized impact of these policies on the
carriers’ prices and market power. Section 2.3 describes the model used in this study.
Section 2.4 presents the results and Section 2.5 concludes. Before proceeding I brieﬂy
summarize my ﬁndings.

First, I discover that reductions in access fees and the development of “equal
access” were responsible for signiﬁcant portions of the decline in interstate MTS rates
and a corresponding reduction in the carriers’ market power.

Second, during this period there were two major changes in the Federal
Communications Commission’s (FCC) regulation of AT&T’S rates for interstate service.
The ﬁrst change was the decision to switch from rate-of-return regulation to price-cap
regulation. The second was the FCC’S reclassiﬁcation of AT&T as a “nondominant”
common carrier, which eliminated price-cap regulatory constraints on AT&T’S domestic

interstate service. As we shall see, neither of these policies were responsible for the

 

market comprised of 1 ﬁrms.

5 See Bresnahan (1989) for a survey of this methodology. The empirical approach to estimating a ﬁrrn’s
market conduct assuming incomplete cost information was classiﬁed by the author as the “New Empirical
Industrial Organization (N 1310).” This title has now become prevalent in the literature.

13

 

 

 

decline
analysi
MTS r:

market

speciﬁc
commOI
explana‘
suscepti
network
conversi

these ex]

difficult

 

decline in MTS rates. In fact, while price-cap regulation had an insigniﬁcant impact, my
analysis demonstrates that rate deregulation actually contributed to the recent increases in
MTS rates and served to restore an appreciable amount of the long-distance carriers’
market power.

Third, while AT&T’s supply relation appears to be robust to the various
speciﬁcations of marginal cost, the supply relation of AT&T’S rivals (referred to as “other
common carriers” or OCCS) is sensitive to the marginal cost speciﬁcations. One possible
explanation for this result is that production of long-distance telephone service is
susceptible to “learning by doing.” That is, (i) the construction of a long-distance
network, (ii) the marketing expertise necessary to acquire a customer base, (iii) the
conversion of customers to equal access service, and (iv) the “learning” acquired from
these experiences introduce an additional component into the cost function that is

difﬁcult to measure or observe.

2.2 Background

A long-distance company operates a communications network that connects the
local exchange carriers’ (LECS’) networks. However, the long-distance network does not
connect directly to its customers. Instead, the LECS, such as the Regional Bell Operating
Companies (RBOCS) and GTE, transport telephone calls between the customers’
premises and the long-distance network. Carrier access to the local network is an integral
input to the production of long-distance telephone service. The price of this input - or
access fee — currently represents approximately 40 percent of the long-distance carriers’

total costs and a substantially greater percentage of their marginal costs.

14

The Modiﬁcation of Final Judgment (MF J) settled the Department of Justice’s
1974 antitrust suit against AT&T and went into effect on January 1, 1984.6 One of the
requirements of the MFJ was that access fees replace the separations and settlements
system.7 The FCC’S subsequent access charge ruling Signiﬁcantly reduced the long-
distance carriers’ burden in recovery of LEC common network costs.8 First, it instituted a
“subscriber line charge” (SLC) which shifted the recovery of a sizable portion of the
LECS’ ﬁxed costs, previously recovered from long-distance companies, to ﬁnal
consumers. The SLC ﬁrst appeared on customers’ local telephone bills in 1984, and the
revenue earned from this charge grew from $1.3 billion to $6.1 billion in 1991.9 Second,
the FCC instituted a series of accounting changes which effectively reduced interstate
costs while increasing intrastate costs. The net effect of these accounting changes was to
reduce carrier access fees an additional $4.5 billion between 1984 and 1991.'0 Figure 2
demonstrates the dramatic overall decline in access fees during the post-divestiture
period. Therefore, it is reasonable to believe that the declining cost of access was
responsible for a signiﬁcant portion of the post-divestiture reduction in MTS rates. The

results strongly conﬁrm this hypothesis.

 

6 The “M” in MFJ refers to the modiﬁcation of a 1956 consent decree to which AT&T had agreed, settling
a 1949 antitrust suit. The MP J required AT&T to divest its RBOCS, thus the terms MF J and divestiture are
often used synonymously.

7 See Brock (1994), p.174. The separations and settlements system, as modiﬁed by the Ozark Plan in 1970,
shifted a disproportionate share of the non—trafﬁc-sensitive costs from local to toll rates. Per Horwitz
(1989, p.354) “The Ozark formula shiﬁed about 1 percent per year of non-trafﬁc sensitive costs to toll
from 1971 to 1984. As a result, the increase in local telephone rates from 1967 to 1984 was the fourth
lowest among all goods surveyed in the Consumer Price Index.”

8 See MTS and WATS Market-Structure, Third Report and Order, CC Dkt. No. 78-72, 93 F.C.C. 2d 241,
251, 252.

9 See Taylor and Taylor (1993). p.186.

'° lbid.

15

Figure 2:
Average Interstate, Real Access Charge for a 10 Minute Telephone Call

1.8

 

1.5

1.2

 

 

 

 

 

 

 

 

 

[blurs
0.9
(8)
0.6
0.3 . . 7777.7 .. 7 77 77 7777 __7__
0 1 I 1 . I I I - 1
1984 1986 1988 1990 1992 1994 1996
YEAR

1+AT81TWADO$ sT—‘Oiarge’ :;;’_OOCWTAooessT—T01arge’ j

16

Another stipulation of the MFJ required the RBOCS to offer all long-distance
carriers “equal access” to the local telephone network.ll Equal access, also called 1+
dialing, allows non-AT&T customers to contact their long-distance carrier without
dialing extra digits. As a result, AT&T’S competitors were able to reduce the amount of
product differentiation in this market by offering services comparable in quality to
AT&T’S. The implementation of equal access by the LECS was an ongoing process
throughout the 1980s (see Figure 3). Today, for all practical purposes, the
implementation of equal access is complete with 99.4 percent of all LEC local lines
providing dialing parity to all long-distance carriers and non-equal access service
accounting for less than 1 percent of non-AT&T long-distance service]2 The hypothesis
of this paper is that the development of equal access service among AT&T’S rivals
diminished the carriers’ market power and in part caused the post-divestiture decline in
interstate MTS prices. The results I obtain conﬁrm this hypothesis.

In July 1989 the FCC ceased regulating AT&T’S interstate rates under rate-of-
return regulation and implemented a more ﬂexible price-cap regulatory policy.'3 The
price-cap plan required AT&T’S rates to fall within a speciﬁc upper and lower price band;
an extended ﬁling period was required for rate requests lying outside of the price band.”

The upper and lower price bands were reduced by 3 percent on an annual basis, of which

 

” See United States v. AT& T, 552 F. Supp. 233.

'2 See FCC (1997a), table 2.3 and 2.20.

'3 See Policy and Rules Concerning Rates for Dominant Carriers, Report and Order and Second Further
Notice of Proposed Rulemaking, CC Dkt. No. 87-313, 4 F.C.C. Red. 2877 (1989).

” The ﬁling process consisted of a submission by the regulated carrier for a new tariff, followed by a 30-60
day period during which the FCC could accept or reject the requested tariff.

17

Figure 3:
The Development of Equal Access

 

80

 

Percent

40

20

 

 

 

 

1984 1966 1988 1990 1992 1994 1996
YEAR

i_._ Local Ines converted to equal access . . .ngShare of OCC equal access mnutes of long-distance service

18

2.5 percentage points represented historical annual productivity growth and 0.5
percentage point was a “Consumer Productivity Dividend.”15

One motivation for implementing price-cap regulation was to enable AT&T to
respond more quickly to ﬂuctuations in the market environment, such as a change in a
competitor’s price. It was argued that removing certain constraints on the interaction
between regulated and non-regulated carriers would enhance competition, promote
efﬁcient ﬁrm behavior and thereby produce a decline in the price of long-distance
service.16 However, Huber, et. al. (1992) and MacAvoy (1996) have argued that granting
AT&T pricing ﬂexibility enables the carrier to retaliate more effectively against a
competitor’s price reduction and that this threat of retaliation, made credible by the
ﬂexibility of price-caps, instead promotes less competition. MacAvoy states “while
[price-cap] tariffs prevented ‘predation,’ they had in place the ﬂexibility to warn the
smaller carriers against strategic discounting to shiﬁ market share.”'7 Huber observed
that since price-caps have been imposed “both Sides have embraced umbrella pricing
under the calming canopy of FCC regulation. AT&T sets the prices. MCI and Sprint
signal their acquiescence in tariffs they insist on ﬁling with the FCC.”18 Therefore, while
a hypothesis of this paper is that price-cap regulation imposed a structural change on the

interaction between long-distance carriers, given the possibility of a competitive or

 

'5 lbid. at l989-97, 2894, 3001.
’6 lbid. at 3066. While efﬁcient ﬁrm behavior was an expected beneﬁt, there was concern with another
possibility: predatory pricing. Speciﬁcally, AT&T might exploit its market power in the provision of
certain services (inbound WATS & MTS) to charge a price below cost for competitive services (outbound
WATS and VNS). See Kahai, et. al. (1995) for a discussion on, and empirical test for, predation in the
long-distance market. Their ﬁndings “yield no support for the argument that reduced regulation has led to
redation. ” (p. 666)
7 See MacAvoy (1996), p.68.
’8 See Huber, et. al. (1992), p.3.44.

19

collusive response, its impact is ex-ante indeterminate. In any case, my results reject the
hypothesis. The estimated impact of price-cap regulation on MTS rates was
economically insubstantial and statistically insigniﬁcant.

In October 1995 AT&T’S unique position as the sole rate regulated interstate
long-distance carrier was eliminated when the FCC unanimously granted the carrier’s
petition to be reclassiﬁed as a “nondominant” common carrier.'9 AT&T not only was
freed from price-cap regulation of its domestic interstate services, but also could ﬁle
tariffs on a one-day notice and did not have to submit cost support data for any proposed
rate change. The FCC’S decision to reclassify AT&T assumed that the carrier lacked
market power in the domestic interstate long-distance market. While the Commission
noted that the market was characterized by a three ﬁrm oligopoly in which AT&T
demonstrated signiﬁcant price leadership, they ruled that the evidence was “conﬂicting

”20 To the extent the condition

and inconclusive as to the issue Of tacit price coordination.
did exist, however, it would be “a problem generic to the interexchange industry and not
speciﬁc to AT&T. ..[and since] they relate to the industry as a whole, these issues do not
preclude our concluding that AT&T lacks the power to raise residential prices unilaterally
above competitive levels.”2|

The impact of rate deregulation also cannot be predicted ex-ante. While complete

pricing ﬂexibility enables the carrier to compete more effectively, it likewise bestows

 

’9 AT&T ﬁled its petition on Sept. 22, 1993: See Motion for Reclassification of A merican Telephone and
Telegraph Company as a Non-Dominant Carrier, CC Dkt. No. 79-252. The petition was ﬁnally granted on
Oct. 13, 1995: See Motion of A T& T Corp. to be Reclassiﬁed as a Non-Dominant Carrier, CC Dkt. No. 95-
427. AT&T’S petition to be reclassiﬁed as a non-dominant carrier in the international long-distance market
was deferred to a later date.

2" See Oct. 13, 1995 Motion.

20

upon the carrier an ability to retaliate against deviations from joint proﬁt maximizing
behavior.22 The results in this paper suggest that rate deregulation discouraged AT&T
from behaving in a more competitive manner. Thus, the FCC’s reclassiﬁcation of AT&T
as a “nondominant” carrier explains a portion of the recent increase in MTS rates and the

restoration of an appreciable amount of market power.

2.3 A Model of Firm Demand and Conduct

First, I describe a static oligopoly model and its speciﬁcation under two
assumptions of marginal cost: (1) complete cost information (i.e., marginal cost is
Observed by the econometrician), and (2) incomplete cost information (i.e., one or more
components of marginal cost are not observed by the econometrician). I conclude the
section by formulating a hypothesis test to determine the marginal cost speciﬁcation that
best ﬁts the data.

Consider a set of two ﬁrms (i = 1,2) each of which supply their service at prices P;

and choose outputs Q; to maximize their proﬁts l'I;,

n1=Pi(QtaQ,/,Yn§t)Qt‘CI(QI,ZI,F71F1) foriij
where, P;( . ) = Firm i’s inverse demand function;
Y; = Vector of exogenous ﬁrm i speciﬁc demand shift variables;
5; = Unknown parameters of ﬁrm i’s inverse demand function;

C;( . ) = Firm i’s total cost function;
Z; = Vector of exogenous ﬁrm i speciﬁc cost shift variables;

 

2‘ lbid. In other words, the FCC passed on addressing a regulatory solution to the collective market power
ossessed by the big three long-distance carriers.

2 Speciﬁcally, cartel stability is facilitated by a member’s ability to inﬂict discomfort upon those that
deviate from the collusive agreement. Accordingly, deregulation may have enhanced the efﬁcacy of
AT&T’S punishment strategies and thus reduced the expected payoff to shirking the cartel agreement.
Under this scenario we would expect to observe an increase (or at least not a decrease) in AT&T’s price.
See Jacquemin and Slade (1989) and references therein.

21

F; = Vector of ﬁrm i’s ﬁxed production costs; and
I", = Unknown parameters of ﬁrm i’s cost function.

The ﬁrst order condition for ﬁrm i with respect to Q; is

d“ - 7P 7P 3C -
——"=Pt+Qtr_l+Qt-(_I_2't_(—'=O (2.1'3)
th 5Q: 5Q; 3Q;
01',
7P 61’» 5C ~
Pi:_Qi ( I "Qi 137+ I (2.1-b)

 

 

 

5Q. 5Q, 0Q:
where, in equation (2.1-a) (P, + Q, 6P, /aQ, + Q, 6P, /aQt, 2,.) represent ﬁrm i’s perceived
marginal revenue, while the ﬁnal term (6C, /aQ,) is ﬁrm i’s marginal cost; equation (2.1-

b) is ﬁrm i’s “supply relation.” The term 71; is a parameter that measures the

competitiveness of ﬁrm i’s conduct and is deﬁned as 6Q, /aQ, .

In the analysis that follows the carriers are aggregated into two groups, where
subscript 1 represents the “dominant” common carrier AT&T, and subscript 2 represents
the “nondominant” or other common carriers (OCCS). There are several justiﬁcations
for aggregating AT&T’S rivals into a Single category. First, for much of the post-
divestiture period there has been asymmetric regulation in the long-distance market.
Speciﬁcally, AT&T has been subject to rate regulation while the OCCS have been rate
deregulated. Second, while the OCCS are quite large in number, they are predominantly
characterized by two ﬁrms with Similar capacity and revenue market shares: Sprint and

MCI.23 Third, there appears to be Signiﬁcant uniformity in the cost functions of the

 

23 Average Capacity Market Share (ACMS): Between 1985-90 MCI and Sprint’s ACMS were 22% and
25%, respectively, and comprised 77% of the OCC ACMS. Between 1991-96 MCI and Sprint’s ACMS
were 23% and 21%, respectively, and comprised 76% of the OCC ACMS. Average Revenue Market Share
(ARMS): Between 1985-90 MCI and Sprint’s ARMS were 10% and 7%, respectively, and comprised 71%

 

 

22

OCCS. Speciﬁcally, access fees are regulated by the FCC. Thus, while there is variation
based on the quality of access (equal vs. non-equal), within a particular category the fee is
uniform across carriers. In addition, non-executive labor is highly unionized in the
telecommunications industry and there is little variation in the union inﬂuenced wage
across carriers.

In the standard empirical application of an NEIO model the ﬁrms’ supply
relations and demand functions are estimated simultaneously. To accomplish this end a
functional form is speciﬁed for each ﬁrm’s inverse demand and marginal cost functions.
In addition, to capture the impact of regulatory change on ﬁrm conduct a function is
speciﬁed indicating the possible determinants of conduct.

I begin with speciﬁcation of the market conduct equation. In the empirical
literature estimates of conduct, such as A; in equation (2.1-b), are used to describe

equilibrium behavior consistent with a particular theory of oligopoly. For example,

,1 ~ an/aQ,
" 037an

for i,j=1,2 and i¢j implies conduct approaching perfect competition;24 21
> 0 and 112 = 0 suggests that ﬁrm 1 is the Stackelberg leader; 711 = 2.2 = 0 is consistent

with Coumot behavior; non-competitive behavior is identiﬁed when 711,712 > 0; and joint

proﬁt maximization is achieved when 2] = ’12 = 1. In the studies by Ward and Kahai, et.
al. the conduct parameters are restricted to predetermined values and assumed time

invariant. In particular, Ward restricts market conduct to be consistent with Coumot

 

of the OCC ARMS. Between 1991-96 MCI and Sprint’s ARMS were 16% and 9%, respectively, and
comprised 65% of the OCC ARMS.

2’ Notice if Q; and Qj are perfect substitutes (6P;/6Q; = 6P;/6Q; = 0) the numerator and denominator of the
expression are zero and the quotient is undeﬁned.

23

behavior and assumes the conduct parameters are constant across carriers and over time:

2;; = 0 for all i and t. Kahai, et. a1. allow conduct to vary across ﬁrms, where the OCCS

are restricted to being price takers while AT&T is the price maker, and assume the

relationship is time invariant: 211; 2 -1 and 212; = -1 for all t. In this paper 1 expand upon
this research by allowing market conduct to vary both across ﬁrms i and over time I. As
speciﬁed below, the model nests a variety of oligopoly models, including those examined
by Ward and Kahai, et.al. In particular, I differentiate between AT&T and OCC market
conduct and examine the changes in the two groups’ conduct that result from regulatory
change. This approach is similar to Spiller and Favaro (1984), Gelfand and Spiller
(1987), and Rubinovitz (1993). The former two papers analyze the market conduct of a
group of “large” and “small” Uruguayan banks and determine the change in conduct
following entry deregulation. The latter paper examines the change in market conduct of
(franchise) monopoly cable TV carriers following rate deregulation.

The linear approximation of ﬁrm i’s market conduct is speciﬁed as

,1, = 2,, + 2,, EA + 71,, PC + 4,, DEREG (2.2)

where, EA equals the percentage of LEC access lines converted to equal access; PC and

DEREG are dummy variables equal to one if AT&T is subject to price-cap regulation or
rate deregulation, respectively, and zero otherwise; 2,0 is ﬁrm i’s (baseline) market
conduct parameter; 2; 1 is the change in ﬁrm i ’5 market conduct parameter attributable to

the development of equal access; and 71; 2 and ,1; 3 measure the change in ﬁrm i’s market
conduct parameter attributed to the implementation of price-cap regulation and rate

deregulation, respectively. Notice that this equation nests the cases where EA, PC, and

24

DEREG have no effect on the ﬁrms’ market conduct and, therefore, allows for time
invariant conduct. Another possible outcome is 21 = ,1 2 and, thus, the speciﬁcation also
nests the case that market conduct is identical across ﬁrms.

Next, I Specify the functional forms imposed on the inverse demand and marginal
cost functions. A log-linear function is assumed for both and speciﬁed as follows:

Inverse demand

 

In E); = 6i0 +6” I“ Q'- +6;2 In QI- +68 In L0CAL+5M In PHONE
' (2.3)
«5,, (1n PHONE)2 + 5,6 In INCOME + 5,, ln QEA, + 6,3TIME + ed,-

The Six exogenous demand shift variables are deﬁned as follows: lnLOCAL is the
natural log of the real consumer price index (CPI) for local telephone service. This
variable is included to control for the hypothesized complementary relationship of local
to long-distance service.25 Therefore, a negative estimated coefﬁcient is expected on
lnLOCAL. lnPHONE is the natural log of the number of US. households that subscribe
to local telephone service, and (lnPHONE)2 is lnPHONE squared. A nonlinear
(quadratic) relationship between lnPHONE and lnP1 is allowed due to the network
characteristics of telecommunications demand: a doubling of subscribers is likely to more
than double demand. lnINCOME is the natural log of real personal disposable income.
An increase in real income should increase the demand for long-distance service. anEA;
is the natural log of the percentage of carrier i’s minutes of interstate long-distance

telephone service that are equal access minutes. Notice that this variable is only pertinent

 

25 For an empirical analysis of the complementary relationship between local and long-distance service see
Hausman, Tardift’, and Belinfante (1993). This hypothesized relationship is also made (and conﬁrmed) in
Ward and Kahai, et. al.

25

Table 2:
Variable Names, Deﬁnitions, and Descriptive Statistics

 

 

 

Variable Deﬁnition Mean Standard Source
Deviation

lnP; Natural log of carrier i’s real tariff price for a ATT = 0.648 0.219
daytime, 100 mile, 10 minute interstate a

Message Toll Service (MTS) telephone call.° OCC = 0.635 0.212

an; Natural log of carrier i’s switched access ATT = 3.849 0.219
minutes for interstate long-distance telephone b

service (in billions). OCC = 3.185 0.623
Q;/Qj Carrier i’s switched access minutes divided by ATT = 2.126 1.017 b

Carrier j’s switched access minutes.
OCC = 0.554 0.193

 

lnLOCAL Natural log of the real consumer price index
(CPI) for local telephone service 4.74 0.060 a
(1982-84=100).*

 

lnPHONE Natural log of the number of US. households

 

 

 

 

 

subscribing to local telephone service 4.49 0.046 b
(in millions).
(lnPHONE? lnPHONE .. lnPHONE 20. 14 0.417 b
lnINCOME Natural log of real per capita personal
disposable income. ’ 9.44 0.059 c
anEAz Natural log of the percentage of OCC interstate
switched minutes that are equal access minutes. 0.901 0.125 b
TIME Linear time trend included to control for

unobserved changes in demand. 23 13.13 —

lnACCESS; Natural log of carrier i’s real average access fee ATT = — 0.46 0.407
for a 10 minute telephone callf a

 

OCC = —0.98 0.489

Sources: (8) FCC, Industry Analysis Division, Reference Book: Rates, Indexes. and Household Expenditure for
Telephone Services (March 1997); (b) FCC, Industry Analysis Division, Trends in Telephone Service (March
1997); and (c) Department of Commerce, Bureau of Economic Analysis. Data are available from their webpage at
http://www.bea.doc.gov.

I MTS is a switched service (i.e., originates and terminates on a LEC’s local network), used primarily by small
businesses and households. The real MTS price is calculated by deﬂating the nominal price with the consumer
price index for all goods and services (i.e., the implicit price deﬂator).

+ Real price are calculated by deﬂating nominal prices with the implicit price deﬂator.

A Access fees are charged on a per minute basis and consist of four components: (I) an originating carrier
common line fee; (2) a terminating carrier common line fee; (3) a trafﬁc sensitive fee; and (4) a non-trafﬁc
sensitive fee. The real access fee is calculated by deﬂating the nominal price with the implicit price deﬂator.
AT&T’S real per minute access fee is based on the fee for equal access. OCCs’ real per minute access fee is
calculated as follows: [%EA x EAFee] + [(1-%EA) x EAFee x (l-%Discount)], where %EA is the percentage of
OCC equal access minutes, EA Fee is the equal access fee, and %Discount is the percentage reduction in access fee
for non-equal access (45%).

 

 

 

26

to the OCCs’ demand function since all of AT&T’s interstate services were offered on an

equal access basis.26 Given the superior quality of equal access service, the hypothesis is

that there is a positive correlation between anEA2 and lnP2 (52, >0). TIME is a linear
time trend included to control for unobserved changes in demand, such as an increase or
decrease in the price of facsimile machines, data processing equipment, etc. No
hypothesis is formulated on the coefﬁcient estimate on TIME.

The assumption of a Cobb-Douglas inverse demand function is consistent with
other empirical studies of the interLATA market.27 This assumption produces the

following partial derivatives with respect to Q; and Q:

i = 571[£] ; and
5Q.- Qt

211-5,,[5]
5Q} Q1"

Marginal cost

 

Complete information: In MC, = 1“,, In ACCESS, + 8,, (2.4-a)
Incomplete information: In MC, = 7,0 + 7,, ln ACCESS, + (15,, (2.4-b)
Inclusion of an intercept parameter (7m) in equation (2.4-b) controls for changes

in the unobserved components of marginal cost. Thus, the exclusion of an intercept in

(2.4-a) presumes marginal cost is comprised of a Single component: ACCESS; The

 

26 The implementation and conversion to equal access technology was a fundamental requirement of the
divestiture. Thus, the development of equal access services among the OCCS is deemed an exogenous
variable dictated by FCC regulation.

2’ See Ward (1995), Gatto, Kelejian, and Stephan (1988), and Gatto, et. al. (1988). The choice ofa log-
linear demand function was supported by results from MacKinnon, White, and Davidson’s (1983) PE test.
While a linear speciﬁcation is rejected for both AT&T and the OCCS inverse demand, the log-linear
speciﬁcation is not rejected for either group (See Appendix 2A).

27

 

assumption that marginal cost is independent of the carrier’s output appears reasonable
given the characteristics of ﬁber optic transmission technology and the signiﬁcant excess
capacity that the facility-based carriers possess. The assumption of a log-linear
functional form is consistent with other empirical studies of the long-distance market.28

The exogenous cost Shift variable lnACCESS; is the natural log of the ﬁrm i
speciﬁc average real price paid to a LEC for a 10 minute telephone call. Access to the
local exchange network is the predominant input required for the production of long-
distance service. Therefore, we would expect an increase in the price of this input to have
a positive impact on the market price.

To derive the supply relation assuming complete cost information, substitute into
equation (2.1-b) the determinants of market conduct as speciﬁed in equation (2.2), the
partial derivatives derived from the inverse demand equation (2.3) and the marginal cost
equation (2.4-a). Taking the natural log of both sides of the equation produces the

following:

In P, = -—ln 1+6” +71,0§,2 21'1'411512 -Q—’EA+71.,26,2 QLPC+71,3§,2 gi—DEREG

Qj Qt Q7 Q1 (25)
+ I“,l ln ACCESS, + 6,,
Equation (2.5) can be Simpliﬁed if one realizes that a linear approximation of the

function y = ln(l + x) 5 x, for the small values ofx.29 Therefore, ﬁrm i’s supply relation

can be rewritten as follows:

 

2“ See Ward (1995).

29 “x” is calculated as the market conduct function 71;, as speciﬁed in equation (2.2), interacted with Q;/Qj
and 5;} Estimates of x, as reported in section V, are the following: i; = — 0.10; 122 = 0.04. A non-linear
regression is performed on equation (2.5) and the results are reported in appendix 2B. The magnitude of

the coefﬁcient estimates is different but the results are unchanged. Given the simpliﬁed interpretation and

28

In P, = —5,, — 2,05,, g—AHEQ —QL EA - 71,275,, —Q—"- PC—7I,35,2 9"— DEREG

.7 Q, Q, Q, (2.6)
+ 1“,, 1n ACCESS, + 5,,

Derivation of the supply relation assuming incomplete cost information is
identical to the derivation of equation (2.6) except the marginal cost function is speciﬁed
according to equation (2.4-b); that is, the intercept term is redeﬁned as — 6;] + 7,-0 instead
of — t5; 1, where the parameter ﬁg represents impact on ﬁrm i ’5 price from changes in the
unobserved components of its marginal costs. Therefore, the two hypothesis are the
following:

H0: Intercept = — 6; 1 [Speciﬁcation (2.4-a)]

H,: Intercept 2 — 6; 1 [Speciﬁcation (2.4-b)]
Speciﬁcally, the alternative hypothesis (H,) is that the intercept equal — 6; 1 + 770.
However, Since 7!;0 is unobserved the inequality condition is sufﬁcient to reject H0. To
determine if speciﬁcation (2.4-a) holds I perform a test of the linear restriction that the
intercept in equation (2.6) equals - 3,, , where 3,, is the coefﬁcient estimate derived from
a regression of the inverse demand equation (2.3).30

The model consists of a system of four time series equations: two equations for
the two ﬁrms, for each time period. The inverse demand equations (2.3) and the supply

relations (2.6) represent the estimating speciﬁcations.

 

 

analysis of the linear model, the results reported in the body of the paper are from the linear regression of
equation (2.6).

30 The test utilizes the system covariance matrix, thus, requiring estimation of the demand and supply
equations simultaneously as a system.

29

2.4 The Empirical Results

The data used are 45 quarterly observations covering the time period 1985:4 to
199624. Given that output and price are determined simultaneously, estimation of the
inverse demand functions and supply relations with ordinary least squares would produce
inconsistent and biased coefﬁcient estimates. Accordingly, this paper utilizes a two-stage
least squares regression procedure. The inverse demand equations contain two
endogenous right hand side variables, an 1 and an2. The observed exogenous cost shift

variables lnA CCESS 1 and lnA C CESS 2 are used as instruments to identify equations (2.3) l

 

and provide consistent estimates of the parameters 6. Likewise, the endogenous right
hand side variable in the supply relation Q;/Qj requires the exogenous demand Shift
variables be used as instruments to identify the supply relations and provide consistent
estimates of the parameters 71.3‘

The inverse demand results are listed in Table 3. The second and third columns
list the coefficient estimates from AT&T and the OCCs’ inverse demand functions,
respectively. Inverting the coefﬁcient on an; provides an estimate of the carriers’ own
price elasticities of demand equal to -1.86 for AT&T and -2.61 for the OCCS. In

comparison to previous research on demand for interstate long-distance service, the

 

3' This identiﬁcation procedure is somewhat atypical for two reasons. First, instead of considering the
correlation between the instruments and outputs Q; and Qj, it examines the correlation between the
instruments and the ratio of ﬁrm i and j’s output (Q;/Qj). Second, the instruments are interacted with the
exogenous variables EA, PC and DEREG. See Appendix 2C for the ﬁrst stage regression results of the
four endogenous variables Q [/Qz, Q2/Q ], an 1, and anz.

30

Table 3:
Two-Stage Least Squares Estimates:
Inverse Demand Functions

 

 

 

 

 

 

Carrier = AT&T OCC
Dependent Variable = lnPl lan
Coefﬁcient Coefﬁcient
Independent Variable (Standard Error) (Standard Error)
INTERCEPT 237.52" 1365.05"
(26.39) (284.52)
an; — 0.538” — 0.383‘
(0.173) (0.212)
1an —— 0.183” - 0.168”
(0.093) (0.075)
TIME 0.014" 0.041”
(0.004) (0.006)
lnINCOME 0.130 0.342
(0.517) (0.480)
lnLOCAL — 0.158” — 0.191
(0.054) (0.158)
lnPHONE -— 572.01 ” —- 607.71 “
(130.06) (127.66)
(lnPHONE)2 63.758" 67.56"
(14.625) (14.37)
anEA 0099'
(0.059)
Observations 45 4S
F-statistic 145.99 138.5
R-Squared 0.9651 0.9760
Adjusted R-Squared 0.9584 0.9696
Durbin-Watson 1.201 1.398
Instruments: lnA CCESS] lnA CCESS]
lnACCESSz lnA CCESSz

 

 

 

Signiﬁcant at the 10 % level
‘ Signiﬁcant at the 5 % level.

31

 

estimates are on the low side but still reasonable.32 The estimated coefﬁcients on the

remaining independent variables in the demand equation are their expected signs, and

most of them are Signiﬁcant. The coefﬁcients on rival output, 612 and 62 2, are — 0.18
and — 0.17, respectively. Given that the numbers are negative and not equal to the
coefﬁcient on an; indicates that rival output is an imperfect substitute for own output.33
The coefﬁcients on TIME are 0.14 for AT&T and 0.04 for the OCCS, suggesting that
unobserved changes in the market, such as increases or decreases in the prices of terminal
equipment have had a positive impact on long-distance demand. The coefﬁcient on
lnLOCAL is — 0.16 for AT&T and — 0.19 for the OCCS. This conﬁrms the hypothesis that
local service is a complement to long-distance service. The coefﬁcient on lnPHONE is —
572.01 for AT&T and —607.71 for the OCCS. Likewise, the coefﬁcient on (lnPHONE)2
is 63.76 for AT&T and 67.56 for the OCCS. This suggests a nonlinear relationship
between subscribership and long-distance demand. The hypothesis of a positive
relationship between subscribership and demand for long-distance service is conﬁrmed
by examining the ﬁrst derivative at the sample means of lnPHONE, PHONE, and P;:
6P1/6PH0NE = 0.05 and 6P2/aPHONE = 0.01.3135 Finally, the positive and signiﬁcant

coefﬁcient on anEA 2 of 0.10 indicates that the development of equal access service

 

32 Ward (1995) estimates a lower bound own price elasticity of demand of 2.02 for AT&T and 3.04 for the
OCCS. Taking these lower bounds as observations (rather than estimates), 1 am unable to reject the

l 1 . .
h th 's that 6 =— t-statistic = 0.34 and 6 = —— t-stat1st1c = 0.56 .
we 88! II 2.02 ( ) 21 3.04 ( )

33 The t-statistics for the hypothesis that the coefﬁcients an1 1 and an12 are identical is 2.89 . Likewise,
the t-statistic for the hypothesis that the coefﬁcient on InQ21 and InQ2 2 are identical is 2.46.

3‘ 6P;/6PHONE = 6,, . (P;/PHONE) + 6,5 .. 2 . (lnPHONE) :1 (P;/PHONE)
6P1/6PHONE = -572.0 .1 (0.022) + 63.8 . 2 . 4.499 4 (0.022) = 0.05
6P2/6PH0NE = -607.7 in (0.021) + 67.6 1.. 2 t 4.499 t (0.021) = 0.01

32

among the OCCS had a positive impact on the demand for their interstate services.36

Table 4 presents the results from estimation of the supply relations. The second
and third columns list the coefﬁcient estimates from AT&T and the OCCS’ supply
relation equations, respectively. The exogenous marginal cost shift variable lnACCESS;
has a positive and Signiﬁcant impact in both equations. The coefﬁcient estimate of 0.83
and 0.67 on lnACCESS; for AT&T and the OCCS support the hypothesis that decreases in
access fees cause signiﬁcant decreases in the market price. While it might appear that
reductions in access fees are being completely passed through to customers in price
reductions (i.e., 1",, z- 1) the t-Statistic for the hypothesis that 1",, = 1 is 2.33 and for the
hypothesis F2, = l is 4.02.

The coefﬁcients of primary interest in this study are those on Q;/Qj, (Q;/Qj)EA,
(Q;/Qj)PC, and (Q;/Qj)DEREG. Notice that the estimated magnitude of the impact of
price-caps on the carriers’ MTS rates is quite small (— 0.002 for AT&T and — 0.06 for the
OCCS) and statistically insigniﬁcant. Therefore, it appears that the change in FCC policy
from rate-of-return to price-cap regulation did not have a structural impact and as a result
market conduct was undisturbed. The estimated effect of a decline in a carrier’s relative
market share on its rates is positive (0.15 for AT&T and 1.07 for the OCCS) and
signiﬁcant. The coefﬁcient estimate on (Q;/Qj)EA is negative (— 0.19 for AT&T and -

1.40 for the OCCS) and Signiﬁcant. These estimates indicates that the development of

 

35 The F-statistic for the hypothesis that lnPHONE and (lnPHONE)2 are jointly signiﬁcant is 3.37 for
AT&T and 4.19 for the OCCS.

36 Recall, AT&T’s service was offered on a 1+ basis throughout the sample period, thus, anEA; is
excluded from AT&T’s ﬁrm speciﬁc inverse demand function. The impact of equal access service on
AT&T’s demand is captured indirectly through the coefﬁcient on an 2.

33

Table 4:
Two-Stage Least Squares Estimates:

 

 

 

 

 

 

Supply Relations
Carrier = AT&T OCC
Dependent Variable = lnP, lan
Coefﬁcient Coefﬁcient
Independent Variable (Standard Error) (Standard Error)
INTERCEPT 0.529W 0.974"
(0.093) (0.133)
Q;/Qj 0.153" 1.071'
(0.021) (0.576)
(Q,/Q,~)EA — 0.192” — 1.397”
(0.038) (0.706)
(Q;/Qj)PC —- 0.002 — 0.059
(0.012) (0.052)
(Q;/Qj)DEREG 0.059“ 0.105“
(0.021) (0.026)
lnACCESS; 0.834” 0674”
(0071) (0.081)
Observations 45 45
F-statistic 582.8 430.96
R-Squared 0.9831 0.9773
Adjusted R-Squared 0.9814 0.9751
Instruments: TIME TIME
lnLOCAL lnLOCAL
lnPHONE lnPHONE
(lnPHONE)2 (lnPHONE)2

 

 

 

Signiﬁcant at the 10 % level
° Signifcant at the 5 % level.

34

 

equal access did affect the carriers’ market conduct. A 10 percent increase in the number
of local lines providing equal access is predicted to have a 3.5 percent reduction in
AT&T’S rates and 6.6 percent reduction in the OCCS rates.37 Finally, the coefﬁcient
estimate on (Q;/Qj)DEREG is positive (0.06 for AT&T and 0.11 for the OCCS) and
signiﬁcant. This is an interesting result that directly conﬂicts with the F CC’s argument
that AT&T no longer possesses signiﬁcant market power. It appears that rate
deregulation affects the carriers’ market conduct by enabling them to exercise more
effectively their market power. Evaluated at the mean of the relative market shares since
rate deregulation was implemented, the estimated impact of this policy is a 7 percent
increase in AT&T’S rates and a 9 percent increase in the OCCS’ rates.

These eight coefﬁcient estimates (as reported in Table 4), along with the estimates
of 6; 1 and c3; 2 (as reported in Table 3), are used to estimate the conduct parameters and

market power. The conduct parameter estimates, evaluated at the mean of EA and

assuming PC, DEREG = 1, are the following:

 

Ii] :2“) +IillEA+112PC+IiI3DEREG
= (0. 15/0.18)—(0.19/0. 18X0.85)—(0.002/0.18)+ (0.06/0. 18) (2.7)
=0.256
Iiz = 220 '1' 1215A + 122 PC + 123 DEREG
=(1.07/0.17)—(1.40/0.17Xo.85)-(0.06/0.17)+(0.11/0.17) (2.8)
= —0.419
37 %AP, /%AEA = 71;;6;2 x Qj/Qj x EA

%AP, /%AEA = -0.19 x 2.13 x 0.85 = .34
(MP2 /%AEA = -1.40 x 0.55 x 0.85 = .65

35

where the standard error for 71, is 0.056 and for 22 is 0.086.38 Notice that, from a purely
computational standpoint, the estimates can easily lie outside the parameter bounds of
perfect competition (A, E -1) and collusion (A, = 1). The apparent reasonableness of the
parameter estimates provides some conﬁdence in the methods used. The economic

interpretation suggest that the OCCS behave in a signiﬁcantly more competitive manner

than AT&T (2, > i, ).39 Given the dramatic shift away from AT&T in output market
share during the sample period this result has some intuitive appeal. However, the results
fail to corroborate any of the predominant theories of market conduct. Joint hypothesis
tests performed on the conduct parameters rejected all of the predominant theoretical
models: perfect competition (21:-611Q1/612Q2 and 22:-621Q2/621Q1), constant

conjectures (A 1=1t2),"0 Stackelberg (21 > O and A 2 = 0), and perfect collusion (71] = 712 =
1).

A second, and closely related, method by which to examine the degree of
competition in the long-distance market is to calculate the carriers’ Lerner indices (4,).
The estimates are derived from supply relation equation (2.1-b) after substituting in the

partial derivatives (w.r.t. Q;) of the inverse demand equation (2.3):

t— P

I

. P-EC/T . - . ,.
,4— 777g! _ ,_ {g}

 

38 Calculation of the standard errors requires estimation of the demand and supply equations as a system.
From this an estimate of the system covariance matrix is produced, and in conjunction with the Delta
method, the standard errors are derived. See Greene (1993), p.297 for a description of the Delta method.

39 The F -Statistic of 8.70 rejects the hypothesis that 71, = 712.
’0 See Bowley (1924). Note that Coumot model is a special case of the constant conjectures model.

36

where, 3,, and 3,2 are the 2SLS estimates, 71, is derived from equations (2.7) and (2.8),

and [—64-] is the mean of ﬁrm i’s relative market share over the entire sample. Imposing
.7
these estimates produces the following Lerner indices:
2, = 0.538 + (O.183)(0.26)(2.l)
= 0.64
2', = 0.383 + (0.168)(— 0.42)(0.55)
= 0.34
where the standard error for .2, is 0.122 and for 22 is 0.064.4|
The ﬁnal result to report is that from the hypothesis test of the marginal cost
speciﬁcations. Recall the two hypothesis are the following:

H0: Intercept = -3,, (complete cost information)

H,: Intercept 2 — 3,, (incomplete cost information)

While the null hypothesis cannot be rejected for AT&T’S supply relation (t-
statistic = 0.1), it is rejected for the OCC supply relation (t-statistic = 2.41). This
suggests that AT&T’S marginal cost is accurately speciﬁed according to equation (2.4-a),
while measurement of OCC marginal cost is more complex and involves additional
components not observed by the econometrician. The intercept parameter estimate in the
OCC supply relation exceeds —3,, (0.974>0.383) suggesting that the net effect of the

unobserved marginal costs components is positive. I provide two possible explanations

 

" Calculation of the standard errors for 2, and .2, follow the same procedure as that for 71, and 712
described in footnote 38.

37

for this outcome. First, aggregation of the OCCs’ supply relations into a Single equation
may be inappropriate. However, this argument is diminished by the apparent uniformity
of marginal costs across carriers. For example, access fees are regulated by the FCC and
variation is primarily attributed to the type of access. Equal and non-equal access fees are
controlled for in calculating lnACCESS 2. In addition, non-executive labor is highly
unionized in the telecommunications industry and there is little variation in the union
inﬂuenced wage across carriers.

A second possible explanation is more compelling and it is the “learning-by-
doing” cost savings that accrue in complex network industries such as long-distance
telecommunications. Constructing a long-distance network from scratch, the acquisition
of marketing expertise,42 upgrading consumers to equal access service, and the
knowledge that is acquired from these experiences may have introduced an additional
component to OCC marginal cost that is not sufﬁciently captured in lnACCESS 2.

To acquire some perspective on the implications of these estimates of market
power it is useﬁrl to make comparisons with other estimates of! in the long-distance
telecommunications market, as well as Similar estimates for other industries. The top
panel in Table 5 includes the results from a variety of empirical studies of market power
as surveyed in Bresnahan (1989). The bottom panel includes the results from Ward
(1995), MacAvoy (1995), Kahai, et. al. (1996) and Knittel (1997). The average market
power index reported in Bresnahan is 0.30. The estimate derived in this paper for

AT&T’S market power (0.64) clearly falls toward the high end of Bresnahan’s reported

 

’2 Recent evidence on Qwest's controversial marketing alliance with US. West and Ameritech suggests
that customer acquisition is a formidable task for new entrants. See Keller (1998).

38

Empirical Estimates of Market Power

Author

Bresnahan’s
Lopez (1984)
Roberts (1984)
Appelbaum (1982)
Appelbaum (1982)
Appelbaum (1982)
Appelbaum (1982)
Porter (1983)
Slade (1987)
Bresnahan (1981)
Suslow (1986)
Spiller-Favaro
Spiller-Favaro

Table 5:

 

Industry

Food processing
Coffee roasting
Rubber
Textile
Electrical machinery
Tobacco
Railroads
Retail gasoline
Automobiles (1 9708)
Aluminum (interwar)
Banks “before”c
Banks “after’”

.4

0.504
0055/0025“
0049“
0.072c
0.198c
0.648c
0.40b
0.10
0.1/0.34d
0.59
0.88/021f
0.40/0.16f

 

Ward (1995)
Kahai, et. al. (1996)
MacAvoy (1995)
Knittel (1997)

Long-distance telephone
Long-distance telephone
Long-distance telephone
Long-distance telephone

0.16/0.07g
0.13h .

0.62/0.61‘
0.65J

 

' Largest and second largest ﬁrm, respectively.
b When cartel was succeeding: 0 in reversionary periods.

C At sample midpoint.

d Varies by type of car; larger in standard, luxury segment.
6 Uruguayan banks before and after entry deregulation.

f Large ﬁrms/small ﬁrms.

g Interstate regional data. AT&T and OCCS respectively.
” Based on AT&T’S output market Share.

’ AT&T and OCCS, respectively.

’ Weighted average based on AT&T, MCI, and Sprint.

39

indices, while the estimate derived for the OCCS’ market power (0.34) is approximately
equal to the average. Given that the industries surveyed in Bresnahan (1989) are likely

biased towards those where the researcher expected to ﬁnd market power, this suggests

that AT&T and, to a lesser extent, the OCCS possess substantial market power.

Second, as indicated in the bottom panel of Table 5, there is Signiﬁcant variation
in AT&T and the OCCS’ estimated Lerner indices. The explanation, argued in this paper,
is that these results are the product of the authors’ overly restrictive assumptions. Recall,
Ward sets the carriers’ conduct parameters equal to zero and assumes they are time
invariant. Kahai, et. al. also assume a time invariant market conduct and restrict AT&T’S
rivals to be a competitive “fringe” subject to a residual demand that is dictated by AT&T.
Finally, MacAvoy and Knittel derive their estimates assuming marginal cost is accurately
measured from accounting data.

This paper overcomes these shortcomings by invoking a model that is sufﬁciently
general. First, unlike Ward and Kahai, et. al., the model does not restrict AT&T’s market
conduct to a particular value but allows it to vary over the sample period subject to
hypothesized regulatory induced changes in the market structure. This speciﬁcation nests
the cases where there is no structural change and no variation in conduct across carriers.
However, this paper discovers that market conduct did indeed vary over time and that the
behavior of AT&T was distinct from that of the OCCS. Second, unlike MacAvoy and
Knittel, the model does not impose a particular speciﬁcation of marginal cost but nests
both the complete and incomplete cost information speciﬁcations and tests which
speciﬁcation best ﬁts the data. The results indicate that AT&T’s supply relation is not

sensitive to either speciﬁcation, while the OCC’S supply relation is indeed sensitive to the

40

Speciﬁcations and is mismeasured under the speciﬁcation imposed by MacAvoy and

Knittel."3

2.5 Conclusion

The F CC’S post-divestiture (de)regulatory policy has produced signiﬁcant gains to
consumers in the form of lower prices. The primary determinants of MTS price decreases
were the development of equal access and the dramatic reductions in access fees. The
ﬁrst results from a change in the carrier’s cost structure and the second from eliminating
an important barrier to entry, namely the technical ability to offer a service of comparable
quality to the incumbent’s service (1+ service).

An important discovery of this paper is that rate deregulation, in particular the
1989 implementation of price-caps and the 1995 reclassiﬁcation of AT&T as
“nondominant”, were n_ot responsible for any of the decline in MTS rates. AT&T and the
OCCS’ MTS rates and Lerner indices of market power actually increased as a result of
this rate deregulatory policy. And the carriers’ estimated 66 and 31 percent price
markups Signiﬁcantly exceeds that which one would expect to ﬁnd from ﬁrms in a
competitive market.

One would hope that the success of the FCC’S equal access mandate, coupled with
the ill effects of rate deregulation, serves as an example to state and federal regulators in

their efforts to deregulate the local telephone market. The current controversy involves

 

’3 OCC marginal cost is underestimated and consequently OCC market power is overestimated.

41

the FCC’S proposed rules in implementing the 1996 Telecommunications Act."4 Those in
opposition to these rules argue that they are too complex and impose regulatory
restrictions which have an asymmetric impact on one class or mode of competitors.45
While this argument would seem to have merit in the abstract, is it valid based on
our past experience in the long-distance telephone market? The development of
competition in the long-distance market has taken well over a decade and has been
assisted by a regulatory environment that has promoted competitive interaction by
removing speciﬁc structural impediments. Were it not for the mandatory deployment of
equal access technology following the AT&T divestiture it is likely that the long-distance
market would not be performing as well as it is. However, the FCC’s equal access
mandate clearly had an asymmetric impact on one class or mode of competitors. In
contrast, FCC policies that removed regulatory asymmetries, such as rate deregulation of
AT&T’S domestic services, enabled AT&T to exercise better its market power and
resulted in a 7 percent increase in the carrier’s MTS rates. According to my analysis of
the long-distance deregulatory experience, to encourage competitive interaction among
the carriers participating in the market (“incumbent” LECS and “competitive” LECS) and
provide incentives for new entry, we need an evolving regulatory policy that addresses

structural imperfections, while constraining the market power of the dominant carrier.

 

4" The 1996 Telecommunications Act was approved February 8, 1996 as PL. 104-104, 1996 (S. 652). The
FCC ’3 rules for implementing competition in the local telephone market were issued August 8, 1996 — see
First Report and Order, In the Matter of Implementation of the Local Competition Provisions in the
Telecommunications Act of I 996 (CC Docket No. 96-98).

45 See Harris and Kraft (1997).

42

Appendix 2A

Residual Demand Speciﬁcation

The PE test, as developed by MacKinnon, et. al. (1983), can be used to test a linear
versus log-linear model speciﬁcation. Let the two competing models be denoted
Linear M0d913 Pi = Bo + 131 Q1 + 132 Qj + 133 Yi + 8:11; and

Log-Linear Model: lnP, = 60 + 6, an, + 82 anJ- + 83 lnY, + ed,

 

Linear Speciﬁcation Test

H03Pi=l30+l31Qi+l32Qj+l33Yi+Sdi
H]: lnP, = 60 + O] ani “l" 52 Ian + 83 lnY, + edl

 

Log-Linear Speciﬁcation Test
H0: lnP, = 60 + O] an, '1' Oz Ian + 83 lnY, + Edi
H13 P1=B0+Bl Qi+l32Qj+B3 Yi+8di

 

Now let b(linear) and d(log-linear) be the two linear least squares estimates of the
parameters vectors. The PE test for H, as an alternative to H0 is carried out by testing the
Signiﬁcance of the coefﬁcient 01 in the following models:

Linear Speciﬁcation Test Log-Linear Speciﬁcation Test

 

 

y = X’B + a {In} — ln(X’b)] + 8 ln y = ln(x)’y + a [( 5, ) _ em)“

] + 8.

The second term is the difference between predictions of y obtained directly from the H,
speciﬁcation (i.e. )3 ) and the log/exponential of the prediction from the H0 speciﬁcation.
The t-statistics presented in Table 6 indicate that the log-linear model m be rejected

in favor of the linear model (t-statistics 0.660 and 0.964). However, the linear model can

be rejected in favor of the log-linear model (t-statistics 1.981 and 2.289).

43

Table 6:
Estimates of 01

 

 

 

 

Estimate Standard Error t-statistic
Linear Test:
AT&T Inverse - 1.721 0.869 1.981
Demand
OCC Inverse — 1.406 0.614 2.289
Demand
Log-Linear
Test:
AT&T Inverse 0.143 0.217 0.660
Demand
OCC Inverse 0.201 0.209 0.964
Demand

 

44

 

Appendix 2B

Nonlinear Regression Results

To examine the appropriateness of the linear approximation of ﬁrm i’s supply
relation equation (2.6), I examine estimates derived from the nonlinear supply relation
equation (2.5). These results are presented in Table 7 and the derivation of the conduct
parameters and Lerner indices follow.

Four coefﬁcient estimates of 2,083, AHSQ, A,25,2, and 236,2 reported in Table 7,
and the estimate of 5,2 reported in Table 3 are used to estimate the conduct parameters.

The parameter estimates are evaluated at the mean of EA = 0.85 and assume PC, DEREG

= 1.
71, = (0.078/O.18)—(0.095/0.18X0.85)+(0.0004/O. 18)+(0.041/0.18) (2.7B)
= 0.215
,1, = (0.250/0.17)—(0.501/0. 17)(0.85)+(0.008/0.17)+ (0.057/0.17) (2.8B)
= —0.652

AT&T and the OCCs’ Lerner indices are derived from the supply relation
equation (2.1-b) after substituting in the partial derivatives (w.r.t. Q,) of the inverse

demand equation (2.3):

1).—ac. a . . ~ - _.
,4, E _'__I/gi_ ,__ -5” .5,2,1, 2'.
P“ Q7

I

where, 3,, and 3,2 are the ZSLS estimates, 71, is derived from equations (2.7B) and

(2.8B) and [gt] is the mean of ﬁrm i’s relative market Share over the entire sample.
1'

45

Table 7:

 

Nonlinear Two-Stage Least Squares Estimates:

 

 

 

 

Supply Relations
Carrier = AT&T OCC
Dependent Variable lnP, lnP2
Coefﬁcient Coefﬁcient
Independent (Standard Error) (Standard Error)
INTERCEPT 0.412" 0.613"
(0.050) (0.043)
Q;/Qj 0.078” 0.250"
(0.010) (0.051)
(Q;/Qj)EA — 0.095" — 0.501 ”
(0.022) (0.090)
(Q;/Qj)PC 0.0004 — 0.008
(0.0046) (0.013)
(Q;/Qj)DEREG 0.041" 0.057”
(0.020) (0.016)
lnACCESS; 0.887” 0.789"
(0.077) (0.085)
Observations 45 45
F-statistic 3746.67 3050.83
R-Squared 0.9989 0.9987
Adjusted R-Squared 0.9987 0.9984
Instruments: TIME TIME
lnLOCAL lnLOCAL
lnPHONE lnPHONE
(lnPHONE)2 (InPHONE)2

 

 

 

Signiﬁcant at the 10 % level
" Signifcant at the 5 % level.

46

 

Imposing these estimates produces the following Lerner indices:

7:, = 0.538 + (0.183)(0.215)(2.1)
= 0.621

2'2 = 0.383 + (0.168)(— 0.652)(0.55)
= 0.323

47

Appendix 2C

First-Stage Regression Results

Table 8:
First-stage Estimates:

Dependent variable InQ,

 

 

 

 

 

Carrier = AT&T OCC
Dependent Variable an, anz
Coefﬁcient Coefﬁcient
Independent (Standard Error) (Standard Error)
INTERCEPT 71.199 548.00"
(64.106) (205.61)
TIME 0.011" 0.006
(0.004) (0.014)
lnINCOME — 0.085 —0.563
(0.524) (0.481)
lnLOCAL —0.65" —1.016’
(0.310) (0.535)
lnPHONE — 126.62” — 239.83"
(56.606) (90.308)
(lnPHONE)2 14.070“ 26.909”
(6.355) (9.929)
anEA —0.158 0.914”
(0.130) (0.433)
InACCESS, —O.461 ” 5.062"
(0.178) (1.334)
lnACCESS2 0.448 —5.437"
(0.401) (1.347)
Observations 45 45
F-statistic 1213.67 550.89
R-Squared 0.9963 0.9742
Adjusted R-Squared 0.9955 0.9694

 

 

 

 

.. Signiﬁcant at the 10 % level
Signiﬁcant at the 5 % level.

48

Table 9:

First-stage Estimates:

 

 

 

 

 

Dependent variable Q,/Q,-
Carrier = AT&T OCC
Dependent Variable Q,/Q2 Qz/Q,
Coefﬁcient Coefﬁcient
Independent (Standard Error) (Standard Error)
INTERCEPT —510.37 585.06'
(450.13) (307.47)
TIME —.0001 0.039”
(0.010) (0.019)
lnINCOME 2.486” —0.438
(1.154) (0.380)
lnLOCAL 3.419” —0.667
(1.690) (0.557)
lnPHONE 791 .66’ -258.12’
(415.46) (136.94)
(lnPHONE)2 —86.807’ 28.769’
(46.691) (15.391)
anEA —3.127” 0.393
(0.935) (0.308)
lnACCESS; —16.433” 2.113"
(2.840) (0.936)
lnACCESS2 17.242” —2.347"
(2.868) (0.945)
Observations 45 45
F-statistic 255.87 144.85
R-Squared 0.9850 0.9739
Adjusted R-Squared 0.9812 0.9671

 

 

 

 

' Signiﬁcant at the 10 % level
Signiﬁcant at the 5 % level.

49

Appendix 2D
Ordinary Least Squares Estimates

In this appendix I present the regression results to the following reduced form
equation:
In P, = a,0 + a,, EA + a,2 PC + (1,, DEREG + a,, In ACCESS, + u,
where, EA equals the percentage of LEC access lines converted to equal access; PC and
DEREG are dummy variables equal to one if AT&T is subject to price-cap regulation or
rate deregulation, respectively, and zero otherwise; and lnACCESS; is the natural log of

carrier i’s real average access fee for a 10 minute telephone call.

Table 10:
Ordinary Least Squares Estimates:

 

 

 

 

Carrier = AT&T OCC
Dependent Variable lnP, lnP2
Coefﬁcient Coefficient
Independent (Standard Error) (Standard Error)
INTERCEPT 1.113" 1.135"
(0.076) (0.091)
EA — 0.822" - 0.414“
(0.092) (0.169)
PC 0.0048 — 0.047
(0.022) (0.212)
DEREG 0.054’ 0.010‘
(0.032) (0.0056)
lnACCESS; 0.760” 0.686"
(0.078) (0.096)
Observations 45 45
F-statistic 665.45 461.01
R-Squared 0.9884 0.9834
Adjusted R-Squared 0.9869 0.9812

 

 

 

TSigniﬁcant at the 10 % level
Signifcant at the 5 % level.

50

.808. $5 8. 008880 €00.50 .0: 0.3 8< 80.80.:0880000... 000. 05 0. 80:883. 300%. c

 

 

 

Ann... 80.803
0.8 :0808 .80
9.68.. Ba 2 2 8:80 am
88088 8.0.8 00 n. .3 0.838500 028:8
05 .0 8038008088. 00.300 2.2.00.8 m. 80.80 :08 :0 80.3.82 832.808.. 0.
98m .800 ”.0250 88808. 2.8800 85.00 05 ha. .003. 80.83 :8 8.8 m...£,..<
.0 80.8.88 80.28 .80.>..0m 0.80.80 889$ 08080.30m .80
.8000 0.8 :0 8.8.820 388.800.8009 00808.0 082? 030. Am. .80 80.8805 05 000.000
b32888. m0DA. .38.. a 8 0:0 8 02.88.00. .80. AM. M80380 8.. 00.... :0... 0. 0008.00. 8820 m800<
08% .808» .80 0.001.. 288.800-8089 8 00 0. 80.500 80.80 8.. 8.8 AN. “80.780 808.80 .30. snow :05 .08.
8.80 80.8088 Hake. 85.80.00. 8:. ..a...< 88.0880 :88 a. 3.8028.— 0: 20.8000 .880: 8 8::
808.800.005.800 .80. 007.. 2.8 £00800 20080.05 00.. £2.0qu 2. ”80.7:3 .80. 8.05 ..0 :0 008qu
3. 8:8. 00 h. 5893. m....om...< ..0 88.80 m 9 000.002 08 m00m.
80.33 00:08... 80.7600. 80.8.28. :0 00808. 0. $5.00 0:... .3006 8.82.800
808.0 8028:. .82 A .. 000—8.. 0058.085. 08.00.... 8.8: 00.... < 80.8.82 8.08000 ..0m. .80. 3.
3 00:80. m. .0< 08880 :25 00.. S... 80.8: .0. 80.. 0880 0800.8 8.8: m88 $830 .030. 80830..
80.8080880020... 88.. 883880 88.. 88.0880 0m... 800.0 088.08. mirage. .05... 00500.). 05 ..0
000. 0:... ”Ea—.50... 00.... “baa—3.. 00... 22:2 00“. “bu—.5; 800... 00... “2.... 80.82.080.08. ”Deanna
awa—
«30. maa— 30. 30. N00. .00. 80— 30. . mum. qua—

 

08.08.... 080382088“.
A: 0.00...

@3300. 0.0.0.53. 9.5382908.— ..0 Dan—8.6

8 808.53.

 

51

Chapter Three

Regulatory Change and Its Impact on Long-Distance Prices
and Market Power: The Intrastate Market

3.1 Introduction

During the years that have followed the AT&T divestiture, the
telecommunications industry has experienced a regulatory environment with a greater
willingness to implement alternative forms of regulation. In July 1989 the Federal
Communications Commission (FCC) ceased regulating AT&T’s interstate services under
rate-of-return (ROR) regulation and implemented a pricing ﬂexibility (PF) policy — price-
cap regulation.I The subsequent, and most recent, change in the rate regulation of
interstate long-distance services occurred in October 1995, when the FCC accepted
AT&T’s request to be reclassiﬁed as a "Non-Dominant" common carrier? 3 The outcome
of this apparently innocuous reclassiﬁcation was the deregulation of domestic rates for

interstate long-distance service.4

 

l Price-cap regulation was approved in May I989 and implemented the following July. A description of
standard price-cap regulation is provided in Section 11. See Kridel, Sappington, and Weisman (1996) for a
description of the various PF regulatory policies.

2 The FCC's post-divestiture policy of regulating long-distance carriers is based on the 1979 Competitive
Carrier rulemaking [77 F.C.C.2d 308 (1979)]. Under Competitive Carrier the FCC concluded that
imposing the regulatory restrictions outlined in Title II of the Communications Act of 1934 on all long-
distance carriers would inhibit, not promote, competition in this market. As a result the FCC established
two categories of carrier: dominant common carrier and non—dominant common carrier. The former,
which consisted of only AT&T, was subject to full FCC rate regulation, requiring a 90 day notice of a tariff
ﬁling. The latter were subject to streamlined FCC rate regulation, requiring a 14 day notice of a tariff
ﬁling.

3 While numerous regulatory changes have occurred subsequent to the FCC ’s reclassiﬁcation of AT&T
(the most publicized being the 1996 Telecommunications Act), no rulemaking or legislation has directly
altered the regulation of interstate long-distance rates.

‘ See Motion of AT&T Corp. to be Reclassiﬁed as a Non-Dominant Carrier, 11 FCC Red 3271, 3288, 1[26
(1995). The FCC still imposes rate averaging based on time of day, distance and duration. In addition,
price-cap regulation is still maintained in the international telecommunications market.

 

52

Table 12:
Recent Changes in the Regulatory Policy of 36 Multi-LATA
States (GA and WV excluded“), 1991-96

 

 

YEAR
REGULATORY
POLICY 1991 1992 1993 1994 1995 1996
ROR Regulation 3 3 1 O O 0
PF Regulation 21 18 15 13 12 l 1
Rate Deregulated 12 15 20 23 24 25

 

 

 

 

* Author was unable to determine the interLATA regulatory policy in GA and WV.

While the decision to implement PF regulation has taken time to develop in the
interstate market, this has not been the case in many of the multi-LATA intrastate
markets.5 In 1988, when the FCC was ﬁrst considering price-cap regulation, 28 of the 38
multi-LATA states had previously eliminated traditional ROR regulation,6 and since then
the intrastate long-distance markets have continued rapidly to reduce and eliminate rate
regulation. For example, by 1991 only three multi-LATA states subjected AT&T to ROR
regulation, and two years later, in 1993, all multi-LATA states had eliminated ROR
regulation. Yet, what is most striking about the contemporary regulatory environment is

the rapid pace at which states have deregulated their long-distance markets. In 1991 , 11

 

5 LATA, short for Local Access and Transport Area, is a geographic entity created at the time of divestiture
as the demarcation boundary for the Regional Bell Operating Companies (RBOCS). Calls which cross the
LATA boundary are referred to as interLATA calls, and must be passed off to a long-distance carrier.
Intrastate interLATA calls are under the regulatory jurisdiction of the state Public Service Commissions
(PSCs). Interstate interLATA calls are regulated at the federal level under the FCC.

6 By January 1988, the following 28 states no longer imposed rate-of-return regulation: Alabama, Arizona,
Florida, Georgia, Illinois, Kansas, Louisiana, Maryland, Michigan, Minnesota, Mississippi, Missouri,
Montana, Nebraska, Nevada, New Jersey, New York, North Carolina, Ohio, Oklahoma, Oregon,
Pennsylvania, South Carolina, Tennessee, Texas, Virginia, Washington, and West Virginia.

53

states had rate deregulated interLATA services;7 by the end of 1996, the number of
deregulated states had increased to 25 (See Table 12).8

It is the transition from a regulated to deregulated market environment that is
examined in this paper. In particular, I employ annual observations on a panel of multi-
LATA states from 1991-96 and compare the market conduct of the long-distance carriers
in regulated markets to that in deregulated markets. My hypothesis is that elimination of
rate regulation altered the carriers' market conduct, and this change in market conduct
explains the price differences in regulated and deregulated markets. More formally, this

is demonstrated by the following generalization of a monopolist's supply relation:9
P = MC - 6QP'(Q) (3.1)
where MC is marginal cost, 6 is the market conduct parameter, and P'(Q) is the slope of

the inverse demand function. The conduct parameter 6 is interpreted as an index
measuring the competitiveness of ﬁrm conduct, where 0 3 HS 1. As Omoves farther

from 0, market conduct moves farther from that of perfect competition. For example, 6=

%V , where N is the number of symmetric ﬁrms in the market, is consistent with Coumot

oligopoly; and (9: 1 is consistent with monopoly.l0
Two procedures are utilized to test the hypothesis of this paper. The ﬁrst employs

the complete time-series of panel data (1991-96). The objective is to estimate a market

 

7 The 11 states deregulated as of December 31, 1991 are the following: Arkansas, Idaho, Illinois, Iowa,
Minnesota, Nebraska, North Dakota, Oklahoma, Oregon, Virginia, and Washington.

8 The 14 states deregulated between 1992-96 are the following: California, Florida, Indiana, Kansas,
Kentucky, Louisiana, Michigan, Missouri, Nevada, New Jersey, Ohio, Pennsylvania, Texas, and
Wisconsin.

9 A supply relation is derived from the ﬁrst-order condition of a ﬁrm’s proﬁt function, with respect to the
speciﬁed choice variable. In equation (1) the choice variable is assumed to be the ﬁrm’s output Q.

54

supply relation similar to equation (3.1). To capture the impact of regulatory change on
market conduct I introduce a linear approximation of 0 that controls for regulatory

change.ll However, to derive a consistent estimate of the change in Hproduced by rate
deregulation I introduce three additional variables that may affect market conduct and be
correlated with regulatory change. Thus, the four determinants of market conduct are
deﬁned as follows: (i) a dummy variable for states that are deregulated (1) or regulated
(0); (ii) a market entry variable measuring the number of facility-based long-distance
carriers competing in a state; (iii) a dummy variable for states that elect (1) or appoint (0)
their commissioners; and (iv) a dummy variable for states that have eliminated (1) or
retained (0) entry barriers into the adjacent intraLA TA toll market. In particular, a single
entry barrier is examined, that is, the technical ability to offer consumers “equal access”
intraLATA toll service.'2

The results indicate that, all things constant, deregulation increased the market
conduct parameter (i.e., reduced the "competitiveness" of market behavior), resulting in a
4.6 percent increase in the market price. Among the political environment variables, the
mandatory provision of "equal access" in the intraLATA toll market had a perverse
impact on the interLATA market, that is, it increased the market conduct parameter,

producing a 5.4 percent increase in the market price. Finally, an increase in the number

 

‘° See Bresnahan (1939), p. 101 8.

11 This approach is similar to Spiller and Favaro (1984), Gelfand and Spiller (1987) and Rubinovitz (1993).
'2 As its title suggests, intraLATA (all service is a long-distance service (i.e., customers are charged a per
minute fee based on distance, duration, and time of day). Following divestiture the provision of
intraLATA services were typically reserved exclusively to the local telephone company. As time
progressed states permitted entry into the intraLATA toll markets, however, only on a non-equal access
basis. It was not until 1995 that a small number of states permitted entry on an equal access basis.

55

Table 13:
A Matrix Description of the Regulatory Experiment Model

 

 

 

 

 

 

 

GROUP
PERIOD Treatment Control
1 Regulated Deregulated
(1991-92)
2 Deregulated Deregulated
(1995-96)

 

of facility-based carriers competing in a market also had a positive impact on the market
conduct parameter, resulting in a 7.83 percent increase in the market price.
The second procedure employs four years of the panel data set: 1991-92 and

1995-96. The objective is to reexamine the linear approximation of market conduct

speciﬁed in the ﬁrst procedure, namely, the assumption that 19 is affected only by rate
deregulation, two political environment variables, and a market entry variable. The
concern is that unobserved factors are partially responsible for the change in market
conduct that results when a state imposes regulatory change. To address this problem I
exploit the regulatory "experiment" conducted by the states between 1991-96. This ﬁrst
entails disaggregating the cross-section of multi-LATA states into two groups: a
"control" group of states deregulated prior to the sample period; and a "treatment" group
of states deregulated between 1993-94. Next, the years 1993-94 are used to separate four
of the six years into two periods: period one (1991-92) is comprised of a deregulated
control group and a regulated treatment group; and period two (1995-96) is comprised of
a deregulated control group and a deregulated treatment group (see Table 13). Based on

this sample I compare the inter-period change in the treatment and control groups' market

56

conduct to uncover and differentiate the inﬂuence of rate deregulation from that of
changes in unobserved factors. Notice that while unobserved factors can, and likely do,
have both a cross-sectional and time-series effect on the model, the procedure just
described focuses only the latter effect. Unobserved cross-sectional variation, referred to
as the state ﬁxed effect, is controlled for by introducing state dummy variables, which I
include in the model. Therefore, the objective of this procedure is to determine if
unobserved changes in market structure occurred between 1993-94 and ascertain how (if
at all) they distorted the estimated change in market conduct attributed to rate
deregulation.

The procedure reveals that unobserved factors had a very small, positive but
statistically signiﬁcant, inﬂuence on the market conduct parameter resulting in a 0.73
percent increase in the treatment group's market price. This suggests the model employed
in the ﬁrst procedure sufﬁciently controls for the impact of structural change on market
conduct. Of equal importance, the results from this "experiment" support the ﬁndings
from the ﬁrst procedure: the deregulatory treatment was determined to cause an increase
in the treatment group's market conduct parameter, resulting in a 6.9 percent increase in
the market price.

The remainder of the paper is organized as follows: Section 3.2 discusses some of
the issues motivating the implementation of alternative regulation in the intrastate
markets; Section 3.3 briefly reviews the literature on this topic and indicates the
advantages of using contemporary data; Section 3.4 ﬁrst describes the general model
employed to estimate the change in market conduct that results from rate deregulation

over the six year sample period. It then presents the regulatory experiment model. This

57

includes a discussion of the criteria for selection of the treatment and control samples, as
well as the methodology employed to distinguish the inﬂuence of the treatment from that
of the unobserved factors; Section 3.5 is a discussion of the results; and section 3.6 states

the conclusions.

3.2 State Motives for Implementing Alternative Regulation

By far the most popular PF regulatory policy implemented at the state level has
been a variant of standard price-cap regulation (See Appendix 3A). Under standard price-
cap regulation a price-ceiling is imposed, typically at the current ROR regulated price and
adjusted periodically, based on changes in exogenous cost factors, such as the carrier's
input costs and productivity improvements. In broad terms, the price-ceiling is set
between regulatory hearings according to the formula:

CAP, =(PP1, — X)CAP,_l

where, CAP, is the price-ceiling, PP], is an index of producer prices, and X is a
predetermined estimate of the change in the carrier's productivity.” Obtaining
government approval to set a price above CAP, involves ﬁling a time-consuming and
expensive rate case. Prices set at or below CAP,, however, are generally effective after a
brief notice period and subject to little regulatory oversight.

An important factor motivating the demise of ROR regulation has been the
dramatic transformation of the telecommunications industry during the post-divestiture

era. For example, the MF J signiﬁcantly altered the ownership structure and the regulated

 

'3 See Kwoka (1993) and Liston (1993).

58

interaction between the carriers in the telecommunications industry. '4 In addition, the
rapid technological change that followed divestiture had a major impact on the method of
long-distance transmission and the cost structure of the long-distance carriers.’5

The response by many state regulators to this dynamic market environment was to
replace the strict constraints of ROR regulation with a PF policy that endowed the
regulated carrier with a (limited) ability to adjust its rates independent of regulatory
approval. One of the rationales for PF regulation was that pricing ﬂexibility would
enable the regulated carrier to respond in a timely manner to unexpected changes in the
market and thus promote efﬁcient regulated ﬁrm behavior. However, it is important to
recognize that reducing regulatory constraints provided only the ability to engage in
competition, not necessarily the incentive to engage in competition. Given the
unfavorable structural characteristics of the long-distance market — large sunk and ﬁxed
costs, a high market concentration, and a dominant ﬁrm controlling well over half the
market — there was little expectation that the incentives to compete would be immediately
forthcoming. Therefore, PF regulation was deemed an intermediate step that (i) enabled
regulators to nurture the incentives of the carriers to compete, while (ii) protecting
consumers from higher prices. It was typically assumed that the equilibrium monopoly

price exceeded the existing ROR regulated price. Consequently, as the incentives to

 

'4 The MFJ had numerous implications for the industry. Some of the more salient aspects include the
following: the divestiture of AT&T's RBOCS, the line of business restrictions imposed on AT&T and the
RBOCS, the required upgrade of RBOC networks to enable all long-distance carriers to offer pre-
subscribed service (also referred to as "1+" or "equal access" service). Coinciding with the implementation
of the MFJ, the FCC reformulated the method by which access fees were paid by the long-distance carriers
for access to the local telephone companies' networks.

'5 In the early 19805, digital ﬁber optic transmission began replacing analog microwave and coaxial cable
transmission. This conversion resulted in a dramatic increase in the capacity of the long-distance network
and contributed to the volatility and variation the competing long-distance carriers' production costs.

59

compete developed under PF regulation one would expect to observe a convergence of
the market price to its competitive level, followed by a transition to a deregulated

environment.

3.3 Literature Review

Mathios and Rogers (M&R) (1989), and Kaestner and Kahn (K&K) (1990)
investigated the impact of alternative forms of state regulation on the price of intrastate
long-distance telephone service between 1986-88. The authors hypothesized that if the
purpose of PF regulation was to restrict AT&T's market power, then the following two
results should be found: (1) in states which had switched from ROR to PF regulation,
there should not be an immediate change in the market price; and (2) in deregulated
states, since AT&T is free to exercise its market power, an increase in the market price
should be observed. The authors examined pricing behavior in three groups of intrastate
interLATA markets: traditional ROR regulated markets, PF regulated markets, and
deregulated markets. M&R employed a single cross-section of data on multi-LATA
states in August 1986 and estimated a reduced form equation that was not explicitly
derived from an underlying structural model of ﬁrm behavior. K&K examined a panel of
multi-LATA states from 1986-88, and estimated a simultaneous system of equations
derived from a speciﬁc market demand function and ﬁrm cost ﬁmction. Both studies
concluded that hypothesis (1) failed to hold. The conversion from ROR to PF regulation
was found to result in a signiﬁcant decrease in the market price.

Interestingly, the research failed to agree on the impact of converting from PF

regulation to rate deregulation. K&K reported that the price of long-distance service in

60

deregulated states was signiﬁcantly lower than in PF regulated states. M&R reported
that, even though the price in both deregulated and PF regulated states was signiﬁcantly
less than the price in ROR regulated states, deregulated states had signiﬁcantly higher
prices than PF regulated states. Although the former research provided further evidence
that removing rate regulation would beneﬁt long-distance consumers, the latter research
contested this conclusion, suggesting that limited constraints on the dominant carrier
might be beneﬁcial to consumer welfare.

It is not entirely clear why the research reaches these opposing conclusions.
While K&K reference M&R in their paper, they fail to reconcile their results with the
earlier work. A possible explanation can be found in the authors’ failure to control
properly for the numerous structural changes that occurred in the years immediately
following divestiture. Two important points follow from this observation. First,
structural change may have produced measurement and speciﬁcation error that is the
cause of the authors’ conﬂicting results. Second, by examining a later time period (1991-
96), the present paper does not expose itself to these difficulties. In particular, the market
structure in recent years has experienced relatively greater stability than in years past. I
provide three examples to support this argument:

(1) Long-distance service has become a fairly homogenous product. Today, all of
AT&T's competitors are provided "equal access" to the local exchange carriers' (LECs)

networks.'6 This development has eliminated an important source of variation in the

 

‘6 Equal access refers to a legal requirement imposed on the local telephone companies to offer comparable
access interconnection to their networks for all long-distance carriers seeking such access by a certain date.
Equal access allowed non-AT&T customers to access their long-distance company without dialing extra
digits and therefore allowed AT&T's competitors to reduce the degree of product differentiation by
introducing services that were comparable to AT&T's.

61

quality of long-distance service. The result is that the price premium for AT&T's service,
which was prevalent throughout the 19808, has today been largely eliminated.l7

(2) The variation in the carriers' costs has diminished signiﬁcantly since the
1980s. While the development of equal access removed much of the product
differentiation in long-distance service, it also eliminated an important cost differential
between carriers. Local telephone companies that persisted in offering non-equal access
were required to charge a discounted access fee for their services ranging from 40-45
percent below the access fee for equal access service. Given that access fees account for
approximately 40 percent of a long-distance carrier's total costs and a signiﬁcantly higher
percentage of its marginal costs, these discounts were an important source of variation in
the carriers' costs of providing long-distance service.'8

(3) Completion of the long-distance carriers’ conversion to ﬁber optic
transmission technology has both reduced the variation in carrier costs and stabilized the
cost structure of the facility-based carriers. In the early 19805, there was a signiﬁcant
change in the production of long-distance telephone service as (low marginal cost) ﬁber
optic technology replaced the pre-existing (high marginal cost) microwave and coaxial
cable technology. The facility-based carriers' investment in ﬁber optic glass was robust
throughout the 1980s, and as a result it increased the variation in costs both across

carriers and over time. For example, conversion of the long-distance network to ﬁber

 

'7 See MacAvoy (1996) and Ward (1996).
'8 Ward (1996) cites access fees at 40% of total cost (p.3), while Kaestner and Kahn (1990) state that
access fees "represent more than 50% of AT&T's marginal costs (p. 366)."

62

optics was not instantaneous but ongoing throughout the 19805. In addition, network
upgrades were undertaken at a different pace by the facility-based carriers. '9

Therefore, for the purposes of this study, the contemporary sample from 1991-96
has advantages over the earlier period examined in M&R and K&K in that it is better
suited to addressing the impact of regulatory change. The homogeneity of long-distance
service and the reduced variation in production costs have diminished the variation in the
carriers' demand and supply behavior. In turn, this reduces the possible sources of
speciﬁcation and measurement error, which may explain the discrepancy in the prior

research results.

3.4 A Model of Market Demand and Conduct
This section consists of two parts. Part A is a description of the general model
employed to estimate the change in market conduct caused by rate deregulation over the

six year sample period. Part B presents the regulatory experiment model.

A. The General Model
To determine the impact of rate deregulation on market conduct of the long-
distance carriers this paper examines a static oligopoly model which does not require the

econometrician to observe all of the components of a ﬁrm's cost function. Instead, a

 

'9 Per the FCC's Fiber Deployment Update (1997) -Table 2, throughout the 19805 AT&T's expansion of its
ﬁber network was dramatic, yet fairly constant. However, in the 19905 there was a signiﬁcant decline in
AT&T‘s investment in ﬁber optics. For example, AT&T added 432,000 ﬁber route miles between 1985-
87, 504,000 ﬁber route miles between 1988-90, 120,000 ﬁber route miles between 1991-93, and 117,000
ﬁber route miles between 1994-96. In contrast, Sprints investment in its ﬁber network was much more
instantaneous and, for the most part, non-existent in the 19905. For example, Sprint added 449,500 ﬁber

63

carrier's marginal cost is inferred from the observed changes in the equilibrium

outcome.20 The model is derived from a ﬁrm's ﬁrst-order condition, or supply relation,

P. = MC.(q.. Z. ,I‘)— M..P.1(Q. Y .6) (3.2)

where, i is the ﬁrm subscript, 5 indicates the market (in our case a state), and t is the

time period. The variables in equation (3.2) are deﬁned as follows:

PM = the market price;

MC,“ (*) = ﬁrm is marginal cost function;

q“ = ﬁrm i's output;

Z, = vector of exogenous cost shift variables;

F = unknown parameters of the marginal cost function;

2 = the ﬁrm speciﬁc market conduct parameter;

st (*) = ﬁrst derivative of the inverse market demand function, with respect
to Q. ;

Q, = aggregate output;

K, = vector of exogenous demand shift variables; and

8 = unknown parameters of the inverse market demand function.

The parameter 1 measures an individual ﬁrm’s market conduct, where ,1 2 0. If A
= 0 price equals marginal cost and conduct is consistent with perfect competition. ,1 = 1
is consistent with Coumot conduct, and joint proﬁt maximizing conduct is identiﬁed
whenever xi. = N, where N is the number of ﬁrms in the industry. Since this paper
employs data at the market level, I proceed by summing equations (3.2) over the N ﬁrms,

yielding

NP... = 2:, MC... (q. Z. ,r) — 1Q.P.1(Q..K. ,5) (3.3)

 

route miles between 1985-87, 17,000 ﬁber route miles between 1988-90, 500 ﬁber route miles between
1991-93, and 1,500 ﬁber route miles between 1994-96.
2° See Bresnahan (1989) for a survey of this "New Empirical Industrial Organization" (N EIO)

methodology.

64

Next, I assume symmetry of the individual ﬁrms' marginal cost functions to

derive the market average supply relation:

P... = MC... (Q. Z .r) — 69.. P. (Q. , Y. ,5) (3.4)
which can be estimated with the industry-level data that are publicly available. Before
proceeding I provide a justiﬁcation for, and an interpretation of, the market supply
relation. The assumption of symmetric marginal costs across carriers can be justiﬁed
given the uniformity of input costs across carriers. For example, the cost of access to the
local exchange network — or “access fee” — is strictly regulated by each states’ PSC.
Variation in access fees across carriers occurs primarily due to the type of access.
However, as of 1991 non-equal access service comprised only a small percentage of the

market.2 In addition, non-executive labor in the telecommunications industry is highly

unionized and there is little variation in the union inﬂuenced wage across carriers. The

interpretation of 6 is the following: 62 %V , therefore 0 S HS 1; perfect competition is

consistent with 6: 0, Coumot behavior is implied when 0: M, , and joint proﬁt

maximizing behavior occurs when 6: 1. Finally, it should not be construed from
equation (3.4) that the ﬁrms necessarily have identical equilibrium conduct. Instead 6
should be treated as the average ﬁrm conduct in an intrastate market.

In the standard application of an NEIO model the supply relation and demand

function are estimated simultaneously. To accomplish this end a functional form is

 

2' For the 36 states examined in this paper the 1991 annual average market share of non-equal access
service was 8 percent [Max = 29.5% (MT-US West), Min = 0.3 (Wl-Ameritech)]. See FCC, Statistics of
Communications Common Carriers (1991-97), Table 2.3.

65

speciﬁed for the inverse market demand and marginal cost function. In addition, to
capture the impact of regulatory change on market conduct a function is speciﬁed
indicating the possible determinants of conduct.

I begin with the speciﬁcation of the market conduct equation. The linear
approximation of market conduct is assumed to take the following form:

6 = 60 + 6l DEREG" + 62 EA." + 63 ELEC 71., + 64 IXCX, (3.5)
where, DEREG“, equals one if state s is deregulated in period t, and zero otherwise; the
ﬁrst political environment dummy variable EA st equals one if state s permits “equal
access” competition in the intraLA TA toll market in period t and 0 otherwise; the second
political environment variable ELECT“ equals one if state s elects its public service
commissioners in period t and zero otherwise; the market entry variable, IXCst, measures

the number of facility-based carriers that provide long-distance service in state s and

period t; 60 is the baseline market conduct parameter, measuring the effectiveness of rate
regulation in inﬂuencing the market outcome; 6, measures the change in market conduct
attributed to rate deregulation; 62 and 6; measure the change in market conduct attributed

to a change in the political environment of the PSC; and 64 measures the change in
market conduct attributed to the entry of a facility-based carrier.
There are no hypotheses imposed on the coefficients on DEREGS, and IXCS).

While rate deregulation and the development of market entry would both be negatively
correlated with the market conduct variable in a non-cooperative oligopoly market (H0261

.<_ 0, 64 S 0), such predictions would not necessarily hold in a collusive market

66

environment. However, with regards to [X C st a positive correlation (H0: 64 > 0) is
suggested by results found in previous research. For example, Kaestner and Kahn (1990)
regress non-AT&T market share on AT&T’s intrastate MTS rates and discover a positive,
statistically signiﬁcant correlation between the variables. In MacAvoy’s (1995) study of
the interstate market, he regresses the weighted average Lerner index of AT&T, MCI, and
Sprint on the Herﬁndahl index of market concentration and discovers a negative
relationship between market power and concentration.22 Therefore, if we assume that

market entry by facility-based carriers produces a signiﬁcant reduction in market
concentration then a ﬁnding of 64 > 0 would be consistent with prior research.

The hypothesis for the coefﬁcient on EA st is 62 > 0. This prediction is based on
the assumption that entry deregulation in the adjacent intraLA TA toll market informs
long-distance carriers of a change in PSC attitude towards (or acquiescence to) market
deregulation. Accordingly, it reduces the probability that regulators will reintervene in
deregulated markets or impose stricter constraints on market conduct in regulated
markets. The end result being that it enables the long-distance carriers to exercise more
effectively market power in their core interLATA market.

Finally, the predicted coefﬁcient on ELECT“ is based on a rent-seeking theory of

23

economic regulation. The hypothesis is that the political support applied by consumer

interests is greater in those states that elect (as opposed to appoint) their commissioners.

Thus, a utility maximizing elected regulator will take greater effort, than an appointed

22 Firm i’s Lerner index .4, s (P i—MC, )/ P, , where P; and MC,’ are ﬁrm i’s price and marginal cost

reSpectively. The Herﬁndahl index is deﬁned as H = 2:4 MS,2 , where MS,- is the market share of the ith

ﬁrm in a market comprised of 1 ﬁrms.

67

regulator, to serve consumer interests. If we assume that regulators inﬂuence market

conduct and consumers favor low rates then the expected coefﬁcient on ELECT“ is 63<0.
Next, I specify the functional forms imposed on the inverse demand and marginal
cost functions. A log-linear function is assumed for both and speciﬁed as follows:

Inverse Demand

 

In P... = 50 + 51 In Q, + 52 In PHONE, + 5, ln INCOME,

(3.6)
+5, ln UNEMPLOYw + 2;” psSTA TE, + 5,,

where, S T A T E S is a dummy variable equal to one if observation 5 is of state 5, and 0
otherwise (W1 is the excluded state dummy). The purpose for including STATES is to
account for possible correlation in the residuals within a state across time, that is the state
ﬁxed effect. The exogenous demand shift variables are deﬁned as follows: lnPHONE is
the natural log of the percentage of total households that subscribe to local telephone
service. An increase in subscribership should increase the demand for long-distance
service. lnINCOME is the natural log of per capita personal income. Given that long-
distance telephone service is considered a normal good, we expect that as a customer's
income rises this will have a positive effect on market demand. anNEMPLOY is the
natural log of the state unemployment rate. As the unemployment rate rises we should
expect the demand for long-distance service to decline. See Table 14 for the summary

statistics of these variables and their data sources.

 

’3 See Stigler (1971), Posner (1974), and Peltzman (1976).

68

The assumption of a log-linear inverse demand function is consistent with other

empirical studies of the interLATA market." Given this functional form, the partial
derivative with respect to Q5, is P; (Q. , K, ,5) = 5. (P, /Q, ).

Marginal Cost

 

1n MC“ = [’0 + I“l 1n ACCESS“ + 50,, (3.7)

The exogenous cost shift variable lnA CCESS st is the natural log of the average
access fee long-distance carriers pay to the Regional Bell Operating Companies (RBOCS)
for a 10 minute telephone call. Access to the local network is the predominant input
required for the production of long-distance service. Access fees generally account for
approximately 40% of a ﬁrm's total costs and an even higher percentage of its marginal
costs. Therefore, we would expect an increase in the price of this input to have a positive
impact on the market price. See Table 14 for the variables, summary statistics, and data
source.

The assumption that marginal cost is independent of the market output appears
reasonable given the characteristics of ﬁber optic transmission technology and the
signiﬁcant excess capacity of the facilities based long-distance carriers.25 The assumption

of a log-linear functional form is consistent with other empirical studies of the interLATA

 

2‘ See Ward (1995), Gatto, Kelejian, and Stephan (1988), and Gatto, et. al. (1988). The choice of a log-
linear functional form was supported by utilizing MacKinnon, White, and Davidson's (1983) PE test.
While the linear speciﬁcation was rejected by both the full sample and the pooled control and treatment
gample, the log-linear speciﬁcation was not rejected by either sample.

As of December 31, 1996 only 53.1% of AT&T ﬁber optic network was "lit," that is, connected to
AT&T switches and capable of transmitting a telephone call. The other facilities based long-distance
carriers have similar excess capacity. See Kraushaar (1997).

69

Table 14:

Variable Names, Deﬁnitions, and Descriptive Statistics

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Deﬁnition Mean Standard Source
Deviation
lnP Natural log of AT&T’s tariffed price for a 0.890 0.200
daytime, 56-124 mile, 10 minute intrastate Control = 0.958 0.210 a
Message Toll Service (MTS) telephone call.‘ Treatment = 0.880 0.207
an Natural log of billed access minutes to the 14.36 1.171
local exchange network (in thousands).+ Control = 13.85 0.989 b
Treatment = 14.87 1.387
lnPHONE Natural log of the percentage of households — 0.065 0.027
subscribing to local telephone service. Control = — 0.058 0.029 c
Treatment = — 0.064 0.023
lnINCOME Natural log of per capita personal income. 9.927 0.256
Control = 9.901 0.240 d
Treatment = 9.933 0.214
In UN- Natural log of the state unemployment rate. — 2.861 0.449
EMPLOY Control = — 2.990 0.488 e
Treatment = — 2.799 0.416
lnACCESS; Natural log of the average access fee long- 0.108 0.396
distance carriers pay to the RBOCs for a 10 Control = 0.21 1 0.434 f
minute telephone call.“ Treatment = — 0.002 0.334
DEREG Dummy variable equal to one if state 5 is 0.551 0.499 g
deregulated in period t, zero otherwise.
EA Dummy variable equal to one if state 5 permits 0.028 0.165 g
“equal access” competition in the intralA TA Control = 0.030 0.173
toll market in period t, zero otherwise. Treatment = 0.030 0.173
ELECT Dummy variable equal to one if state 5 elects 0.222 0.417 g
its public service commissioners in period t, Control = 0.273 0.449
zero if commissioners are appointed. Treatment = 0.091 0.289
IXC The number of facility-based carriers that offer 6.379 2.665 g
long-distance service. Control = 6.969 2.817
Treatment = 6.671 2.582

Sources: (:1) Tariffed MTS rates are ﬁled with the state Public Service Commissions; (b) FCC, ARMIS 4308 —
Annual Operating Statistics and Access Line Data (annual reports from 1991-96), Table IV; (c) FCC, Industry
Analysis Division, Telephone Subscribers/tip in the US. (1997); (d) US. Department of Commerce, Bureau of
Economic Analysis, State Personal Income Report; (e) US. Department of Labor, Bureau of Labor Statistics; (1)
FCC, ARMIS 4308 and ARMIS 430] — Quarterly Financial and Demand Data (4th quarter reports from 1991-96);
and (g) NARUC, Utility Regulatory Policy in the U. S. and Canada (1990-96) and phone survey of commissions.
Notes: * MTS is a switched service (i.e., originates and terminates on a LEC’s local network), used primarily by
small businesses and households. The mileage band range 56-124 was chosen because it was the most popular

mileage band used by state PSC5(14 of 36). For states that classiﬁed MTS prices according to a different mileage
band, a weighted price was calculated that corresponded to a 56-124 mile call. See Mathios and Rogers (1989) for
a similar treatment of mileage band variation.

4' Billed access minutes (BAMs) are compiled by the local telephone companies and are based on the bills they
issue to the long-distance carriers for access to the local network. BAMs only include minutes of switched
interLATA service on a minute-of-use basis. In particular, BAMs do not include special access services (i.e.,
“service bypass”) which employ dedicated circuits leased from the local telephone companies. BAMs also do not
include long-distance service provided over leased or purchased facilities not owned by the local telephone
company (i.e., “facilities bypass”).

A Access fees are charged on a per minute basis and consist of four components: (1) an originating carrier
common line fee; (2) a terminating carrier common line fee; (3) a traffic sensitive fee; and (4) a non-trafﬁc
sensitive fee.

70

market.26

To derive the market supply relation substitute into equation (3.4) the
determinants of market conduct as speciﬁed in equation (3.5), the partial derivative from
the inverse demand equation (3.6) and the marginal cost equation (3.7). As in the inverse
demand function, state dummy variables are included to control for possible correlation
in the residuals across states. Taking the natural log of both sides of the equation

produces the following:

ln P, = — 1n[1 + (5,5, + 0,5, DEREG, + 5,5, EA, + 5,5, ELECT, + 5,5, LXC, )] 3 8
+r0 + r, In ACCESS, + 2:,” AS“ TE, + 5., ( ' )
Equation (3.8) can be further simpliﬁed by taking a linear approximation of the
natural log function. Note that y = ln[1+ x] s x, for the small values of x which are found

in this study.27 Imposing this approximation removes the non-linearity of equation (3.8)

and results in the following:

In Psi = _60(sl — 01(51DEREGM — 0261EA." .— @dELECT'" — 6461MCM

S=35 (3’9)
+1“0 + r, In ACCESS, + Z... ,6,STA TE, + 5,

WI
Therefore, the general model consists of a system of two equations as speciﬁed by

the inverse market demand (3.6) and the market supply relation (3.9).

 

26 In Ward's (1995) study of the intrastate interLATA market he estimated a log-linear supply relation that
imposed a log-linear marginal cost function. The dependent variable in the supply relation was an index of
the price of MTS and the estimated coefﬁcient on output in the supply relation was insigniﬁcant. In
addition, Kaestner and Kahn's (1990) supply relation was log-linear, while Mathios and Rogers' (1989)
reduced form equation was also log-linear.

27 A non-linear regression was performed on equation (3.8) and the results reported in appendix 3B. The
magnitude of the coefﬁcients estimates are similar and the results are unchanged. Given the simpliﬁed
interpretation and analysis of the linear model, the results reported in the body of the paper are from the
linear regression of equation (3.9).

71

B. The Regulatory Experiment Model

Notice equation (3.5) assumes that only four variables inﬂuence market conduct:
rate deregulation, two political environment variables, and a market entry variable. In
this section a unique approach is taken to determine whether the linear approximation of
market conduct is properly speciﬁed. The objective is to address the following questions:
Are there additional, possibly unobserved, variables that inﬂuence market conduct and
are excluded from equation (3.5)? And if so, how does their inclusion alter the results
previously derived?

Three examples of unobserved factors are provided. First, even at the present
time no state has completely deregulated intrastate long-distance service. Generally, the
states that have ceased regulating rates still have an inﬂuence on the market environment
by requiring statewide uniform pricing (i.e., no geographic de-averaging) and imposing
limited reporting requirements. This suggests that while regulatory oversight plays an
explicit role in regulated markets, as reﬂected in the regulatory statutes, it appears also to
play an implicit role in both regulated and deregulated markets, as reﬂected by the
attitudes of state regulators and their interpretations of the market environment. Changes
in these attitudes and interpretations could alter market conduct. Unfortunately, such
changes are difﬁcult to measure and often unobserved. Second, during the sample period
there have been numerous actual, and anticipated, changes in FCC and state regulation

that indirectly affected the intrastate long-distance market.28 This deals with the inﬂuence

 

2’ For example, the FCC deregulated interstate long-distance services to high volume business consumers
in October 1992, and deregulated interstate 800 services (excluding 800 directory services) in May 1993.
In 1994 Congress began deliberations in what would eventually result in the 1996 Telecommunications
Act. Finally, in 1994 a small number of states began examining whether to permit 1+ competition in their
intraLATA toll markets.

72

of multi-market contact on market conduct, which one could argue is partially controlled
for in the model by EAst. However, given the multitude of markets in which the long-
distance carriers compete EAst may be an insuﬁicient control for multi-market contact.
Finally, the ﬂow of information between the carriers may have improved during the
sample period as a result of the convergence and stabilization of the carriers' production
costs in the late 19805. A change in the ﬂow of information between carriers could have
an inﬂuence on their market conduct.

In deriving the regulatory experiment model I begin by disaggregating the cross-
section of states into two groups: a "control" group of states deregulated prior to the
sample period; and a "treatment" group of states deregulated in 1993 or 1994. Next, the
years 1993-94 are used to separate the panel data set into two periods: period one (1991-
92) is comprised of a deregulated control group and a regulated treatment group; and
period two (1995-96) is comprised of a deregulated control group and a deregulated
treatment group (see Table 13).

The purpose for deﬁning two time periods is to compare the inter-period change
in the treatment and control group's market conduct. Since the control group has
previously experienced rate deregulation, the inter-period change in this group’s market
conduct, controlling for ELECT, EA, and IXC, must be attributed to unobserved factors.
However, the inter-period change in the treatment group's market conduct is subject to
both the deregulatory treatment and unobserved factors. To utilize the information
provided by the control group, the impact of unobserved factors on market conduct is
assumed constant across the groups and uncorrelated with the treatment. From this a

comparison of the inter-period change in market conduct in control and treatment groups

73

uncovers and differentiates the inﬂuence of rate deregulation from that of the unobserved
effects.

The sample employed in the regulatory experiment model differs in two respects
from that utilized in the general model. First, as previously described, the regulatory
experiment excludes two of the six years of the panel. Such an adjustment is necessary to
acquire a treatment group of sufficient size; that is, the compilation of a time-series of
treatment states that corresponded with the time-series of the control states required
expanding the inter-period interval to two years. Second, the cross section of states in the
regulatory experiment is a subset of the cross-section utilized in the general model. This
occurs because not all of the multi-LATA states qualiﬁed for inclusion in the
“experiment.” For example, 10 states that retained regulation for the duration of the
sample period were excluded.29 In addition, 4 states that were deregulated during the
sample period but not at the speciﬁed inter-period interval were also excluded.30
Therefore, the treatment group consists of annual observations on 11 states for four years
- 1991, 1992, 1995, and 1996;3| the control group, coincidentally, also consists of 11
states and the same four years.32

The linear approximation of market conduct required a slight modiﬁcation of

equation (3.5) and is speciﬁed as

 

29 The 10 states that retained regulation for the duration of the sample period are the following: Arizona,
Colorado, Maryland, Massachusetts, Mississippi, Montana, New York, North Carolina, South Carolina,
and Tennessee.

30 Three states were deregulated in 1992: Indiana, Michigan, and New Jersey. Alabama was reregulated in
1 995.

3 I The treatment group of states deregulated between 1993-94 are the following: California, Florida,
Kansas, Kentucky, Louisiana, Missouri, Nevada, Ohio, Pennsylvania, Texas, and Wisconsin.

32 The control group of states deregulated prior to 1991 are the following: Arkansas, Idaho, Illinois, Iowa,
Minnesota, Nebraska, North Dakota, Oklahoma, Oregon, Virginia, and Washington.

74

5 = 50 + (5, + 5, )TIME, + 52 EA, + 5, ELECT, + 5, IXC, (3.10)

where a, is the change in market conduct attributed to unobserved factors, and TIME, is
a dummy variable equal to one for period two observations and zero for period one

observations. The coefﬁcient on TIME, states that market conduct in period two is
affected by deregulation of the treatment group (6,) as well as a change in unobserved

factors ((1,). Recognize that by deﬁnition 6, = 0 for the control group, since it
experienced no inter-period change in rate regulation.

For efﬁciency reasons, the treatment and control samples are combined and
estimated in a single regression equation. In addition, a number of linear restrictions are
tested to determine if the supply and demand coefﬁcients are equal across groups. In
Section V I discuss the restricted regression results. However, in the following
discussion I present the unrestricted model and list each group's demand and supply
relation equations separately.

The control and treatment supply relations are identical to equation (3.9), except
that market conduct is respeciﬁed according to equation (3.10). The supply relations are
listed below. Note that superscripts are added to distinguish between the treatment (T)
and control (C) groups.

Control Group

 

In Psi : —60('51(‘ _ a161(TIMEt _ 62"6ICEA,’ - 63(61CELECTH _ 64('5l('I/YC,,
. . H, ,. (3.11-C)
+1"; + r,‘ In ACCESS, + 2,, 5, STA TE, + 5,

75

Treatment Group

 

In P. = -5357' — (6.” + a. )6." TIME. - 5,155.4, - 575751.555, — 61mm.
(3.11-T)

S=10

+r," + r," In ACCESS, + Z ,5," STA TE, + 5,

5:1 .,.,
By failing to include a group superscript on a, I impose the assumption that
unobserved effects are identical across groups. In the deregulated control group, 6," is by
deﬁnition zero and dropped from the control group supply relation. Thus, the coefﬁcient
on TIME, in equation (3.11-C), alcfl" , measures the change in market conduct between
periods one and two attributed to a change in unobserved factors (a,) interacted with the

control group's demand elasticity (6," ). The treatment group is subject to both rate

deregulation and unobserved effects and, accordingly, the coefﬁcient on TIME, in
equation (3.1 l-T), (W + a, )6)”, measures the change in market conduct between periods
one and two attributed to deregulation (6,") and unobserved factors (a,), interacted with

the treatment group's demand elasticity (6," ).

The inverse demand equations are identical to equation (3.4) except that
superscripts are added to the coefﬁcients to distinguish between the treatment(T) and
control(C) groups.

Control Grog;

 

In P, = 5," + 6,‘ ln Q, + 5," 1n PHONE, + 5; ln INCOME,
5.10 5 (3.12-C)
+5: 1n UNEMPLOY, + Z... ,1, STA TE, + 5

(is!

76

Treatment Group

 

1n P, = 5,,T + 5,T In Q, + 5,T ln PHONE, + 5," ln INCOME,

, , (3.12-T)
+5] ln UNEMPLOY, + 2’,” ,1] STA TE, + 5

(1.51
The hypothesis I wish to test is whether rate regulation restricted market conduct.

The formal statement of this hypothesis is the following:
H0: 6," > 0
Hl: 6," S 0

For H0 to be valid, there must be a positive inter-period change in the treatment
group's market conduct attributed to rate deregulation. However, interest lies not only in
the sign of this expression, but also its magnitude and the its subsequent impact on the
market price. These issues are discussed below.

First, I examine the coefﬁcient estimate on TIME, in the treatment group's supply

'I'

relation equation (3.11-T): (6," + Oil )5 .33 An estimate of demand elasticity 3," is

attained from the inverse market demand function, and

T

' (3.13)

 

 

6T 6
61T+al=(l gial)
1

However, from the estimate derived in (3.13) one cannot distinguish 6,]. from 5,.

To address this problem I take advantage of the information provided by the control

group plus the assumption that a, is constant across groups. The coefﬁcient on TIME, in

equation (3.11-C) is the following: Ki. An estimate of the demand elasticity 3:? is

 

33 The bar over the variable and those that follow indicate an estimate.

77

attained from the control group's inverse market demand function and, accordingly, a, is

estimated as follows:

_ 616“
l I
a = — ' 3.14

 

Finally, a "difference-in-difference" estimator” is employed to uncover the

estimate of 6,7, which is calculated by taking the difference of the treatment and control

groups' inter-period differences in market conduct,
5." =5,T—+5,l —E. (3.15)
Derivation of the standard errors of the estimates derived in equations (3.13)-
(3.15) requires estimation of the demand and supply equations as a system. From this an

estimate of the system covariance matrix is produced, and in conjunction with the Delta

method, the standard errors are derived.35

3.5 The Empirical Results

Six years of data are collected on 36 of the 38 multi-LATA states at annual
frequencies from 1991 to 1996.36 The general model utilizes all 216 observations of the
panel data (T=6, S=36) and is comprised of two equations: (3.6) and (3.9). The

regulatory experiment model employs four years of data on a cross-section of 22 states,

 

3’ The application of a "difference-in-difference" estimator to uncover variation between a treatment and
control sample has typically been utilized in the labor economics literature. See Gruber (1994), Card and
Krueger (1995), and Madrian (1994).

35 See Greene (1993), p.297 for a description of the Delta method.

36 Two states, Georgia and West Virginia, were excluded from the sample since the author was unable to
determine these states' long-distance regulatory policy.

78

or 88 observations (T=4, S=22), and is comprised of four equations: (3.11-C), (3.12-C),
(3.1 l-T), and (3.12-T).37

Table 15 presents the results from estimation of the log-linear demand equations.
Column two lists the demand results from the general model. Columns three and four list
the restricted demand results from the regulatory experiment model. Where a single
number is listed in the third and fourth column, the variable is restricted to having the
same impact across groups.38 Given that the market price and output are determined
simultaneously, estimation of the inverse demand functions with ordinary least squares
would produce inconsistent and biased coefﬁcient estimates. Accordingly, I utilize a

two-stage least squares regression procedure. The primary purpose for regressing

demand is to obtain estimates of the coefﬁcients on InQ,: 61,5? and 3:. These
coefﬁcients are then used to estimate the determinants of market conduct and the
unobserved effect.

Inverting the coefﬁcient on an, provides a measure of the price elasticity of
demand. The estimated elasticities are -0.742 for the full sample regression, -0.835 for
the control group, and -0.654 for the treatment group. In light of previous empirical

research on demand for long-distance service, the estimated elasticities are reasonable.”

 

37 Note that for estimation purposes the four treatment and control equations are combined into two
equations [(3.9-C)+(3.9-T)=(3.9) and (3.10-C)+(3.10-T)=(3.10)]. The two equations are estimated with
the pooled sample consisting of both groups.

38 In the ﬁrst “stage” I estimated the unrestricted inverse demand function and performed joint hypothesis
tests on each pair of treatment and control coefﬁcients. In the second “stage” I estimated the restricted
inverse demand function, where the paired coefﬁcients that failed to reject equality across groups were
restricted to a single coefﬁcient estimate. For the paired coefﬁcients that rejected equality across groups no
restrictions were placed on the coefﬁcients.

39 See Ward (1995), p.38, Kahai, et. al. (1996), p.509 and Taylor (1994).

79

 

Table 15
Two Stage Least Squares Estimates: Inverse Demand Function
Dependent Variable: lnP,
(t-statistics in parentheses)

 

 

 

 

 

 

The General The Regulatory
Model Experiment
Control Model Treatment
Full Sample Group* GroupM
(S=36, T=6) (S=11, T=4) (S=11,
T=4)
INTERCEPT 2.291 .2704
(.503) (.0826)
an -1.348 -1.197 -1.528
(-2.664) (-2.131) (-2.205)
lnPHONE -.177 1.120
(-.195) (.922)
lnINCOME 1.226 1.574
(18.512) (2.731)
anNEMPLOY -.1592 -. 1269
(-2.47) (-1 .993)
STATE
DUMMY YES YES
VARIABLES
F-statistic 1048.58 2186.22
R-Squared .9957 .9962
Adj. R-Squared .9948 .9958
Instrument: lnA CCESS lnACCESS

 

*Control Group States —

1) AR, 2) ID, 3) IL, 4) IA, 5) MN, 6) NE, 7) ND, 8) OK, 9) OR, 10) VA, and 11) WA.
“Treatment Group States -—
1) CA, 2) FL, 3) KA, 4) KY, 5) LA, 6) MT, 7) NV, 8) OH, 9) PA, 10) TX, and 11) W1.

80

 

The estimated coefﬁcients on the remaining independent variables in the demand
equations, with one exception, have their expected signs and are signiﬁcant. The
coefﬁcient estimate on lnINCOME is 1.23 for the full sample regression. In the pooled
control and treatment regression the restricted coefﬁcient estimate is 1.57. These results
conform with the expected income effect on a normal good. The estimated coefﬁcient on
state unemployment rate In UNEMPLOY is -.159 for the full sample and for the pooled
treatment and control sample it is -.127. These results agree with the hypothesized
impact of unemployment on demand. Finally, the variable measuring the number of
households that subscribe to local service lnPHONE has a coefﬁcient estimate of -. 1 77
for the full sample, and for the pooled sample the restricted coefﬁcient estimate is 1.12.
Neither of these point estimates are statistically signiﬁcant. A more broadly deﬁned
statistic is collected by the FCC measuring households with access to telephone service.
Similar results were found when this measure was used in place of lnPHONE. For all
states in the sample lnPHONE was stable (stand. deviation = 0.024) with a slight upward
trend (average annual percentage change = 0.1) and in only three states did it drop below
90 percent (AR, AL, MS). This suggests that a possible explanation may be the success
of "universal service," that is, the variation in lnPHONE, across states and over time,
represents primarily transitory shocks which do not signiﬁcantly affect demand.

Table 16 presents the results from estimation of the supply relation equations.
Column two lists the supply relation results from the general model. Columns three and
four list the restricted supply relation results from the regulatory experiment model.

Where a single number is listed in the third and fourth column, the variable is restricted to

81

Table 16

Two Stage Least Squares Estimates: Supply Relation
Dependent Variable: lnP,
(t-statistics in parentheses)

 

 

 

 

The General The Regulatory
Model Experiment
Control Model Treatment
Full Sample Group Group
(S=36, T=6) (S=11, (S=11, T=4)
T=4)
INTERCEPT .5363 .9057
(3 .936) (3 .572)
DEREG .0463
(2.664)
TIME .0057 .0766
(1.881) (3.955)
ONEPL US .0535 .1248
(2.693) (4.328)
ELECT -.2208 -. 1092
(-.942) (-1 .73 8)
IXC .0783 .1546 .1007
(2.490) (3 .001) (2.466)
lnACCESS .2475 .3430
(6.368) (6.840)
STATE
DUMMY YES YES
VARIABLES
F-statistic 49.23 60.52
R-Squared .9161 .9402
Adj. R-Squared .8970 .9546
Instruments: lnINCOME lnINCOME
anNEMPLOY In UNEMPLOY

 

 

 

82

 

having the same impact across groups. Given the endogenous right-hand side variable
IX C 5,, I use two-stage least squares regression to estimate the supply relation equations.
The coefﬁcients of primary interest are on the dummy variables DEREG and
TIME. The estimated coefﬁcient on DEREG indicates an increase in the market conduct
parameter following deregulation, resulting in a 4.6 percent increase in the market price.
For the control and treatment groups the estimated coefﬁcient on TIME indicates that the
inter-period change in market conduct was responsible for a 0.57 and 7.7 percent increase

in the market price, respectively.

To uncover an estimate of 6,, a,, and (6,T+ a,) it is necessary to divide the
coefﬁcients on DEREG and TIME by the estimated coefﬁcients on an as listed in Table

15. Standard errors and t-statistics are derived by employing the Delta method. For the

full sample of 36 states 61 = 0.034 (s.e.=0.018), for the control sample 3, = 0.0048 (s.e.=

 

0.0034) , and for the treatment sample 6] + al = 0.050 (s.e.= 0.026).
The "difference-in-difference" estimator is employed to uncover an estimate of

6,]. as follows:

 

5," +5, —3. = .0501 - .00476

‘5'." .0453

The standard error of 6i is 0.021. Therefore, the estimated change in market conduct

attributed to deregulation (6? =0.045) is responsible for a (.04SXI .528) 6.92 percent
increase in the market price (s.e.=.0301). See Table 17 for a summary of the estimated

market conduct results.

83

(t-statistics in parentheses)

Table 17
Summary of Results

 

 

 

 

 

 

 

The General The Regulatory
Model Experiment Model
Full Sample Treatment Group
(S=36, T=6) (S=11, T=4)
A9 => APS, AGT 3 APSI
Intrastate Deregulation 3.43 % 4.63 % 4.53 % 6.92 %
(0,) (1.905) (2.664) (2.191) (2.296)
Unobserved Effects 0.48 0.73
(m) (1.511) (1.201)
Equal Access 3.98 5.35 8.17 12.48
IntraLATA Toll Comp. (1.913) (2.693) (1.965) (4.328)
(92)
Elected State Public -16.38 -22.08 -7.15 -10.92
Service Commissioners (.8912) (-.942) (1.465) (-1.738)
(93)
Facility-Based 5.81 7.83 6.59 14.07
Interexchange Carriers (1.824) (1.490) (1.647) (2.466)

( 94)

 

 

 

 

Both the results from the general model, employing the full sample of 36 states,
and the results from the regulatory experiment model, employing the pooled control and
treatment samples, support the hypothesis that regulation restricted the market conduct of
the long-distance carriers. However, notice that one interpretation of this result is that
regulators in price-cap states set price ceilings below the average costs of production.
Therefore, the estimated increase in the market conduct parameter, which in turn caused
an increase in the market price, instead reﬂects the convergence of the market price to a
sustainable level and not the exercise of market power. Fortunately, the behavior of the
market price under price-cap regulation makes this interpretation highly unlikely. M&R
indicate that in more than half of the 9 states that imposed price ceilings in 1986, AT&T's
rates were signiﬁcantly below the ceiling. In addition, Ward's (1995) study of the
interstate market in the 1990s found that price ceilings were binding on AT&T's rates for
less than one-third of the price-cap period and Ward argues that even this measure may be
overstated. Since the cap was more oﬁen binding after large changes due to signiﬁcant
decreases in the regulated carrier access prices, this one-third may reﬂect regulatory delay
in reviewing AT&T rate changes.

The estimated impact of the political environment and market entry variables
provide some interesting results. In both regressions the estimated coefﬁcient on EA was
positive and signiﬁcant. As presented in Table 6, when a state introduced equal access
competition in the intraLA TA toll market this produced a 5.4 percent increase in the
market price for the general model and 12.5 percent increase for the regulatory
experiment model. This suggests that competition between the long-distance carriers

diminished when entry to the intraLATA toll market was liberalized. It is important to

85

recognize that this result holds regardless of the regulatory policy in the interLATA
market, suggesting that implicit constraints on the carrier’s rate setting behavior remain
even after explicit rate regulation is eliminated. Therefore, changes in regulatory policy,
whether applicable to the long-distance market or an external market, provide information
to the long-distance carriers on the commitment of regulators to a deregulated market
outcome. Accordingly, entry deregulation to the intraLATA toll market signals a pro-
market attitude, which reduces the carriers’ perceived probability of regulatory
reintervention and, in turn, enables them to exploit more effectively their market power.

The coefﬁcient estimate on ELECT provides some support for the hypothesized
effect. In the pooled control and treatment regression, electing a public service
commissioner had a statistically signiﬁcant effect on conduct of -10.92 percent (t-stat=-
1.738), whereas in the full sample regression the point estimate is -22.08 percent but is
not statistically signiﬁcant (t-stat=-0.942). The expected negative impact is based on a
rent-seeking theory of regulation, that is, when a regulator's job is more closely linked to
serving consumer interests, she will take more effort to restrict the conduct of the ﬁrms
participating in the market.

The estimated coefﬁcient on IXC is positive and signiﬁcant in both regressions,
indicating that the market price is higher in states where there are more facility-based
carriers; or an increases in the number of carriers over time has a positive impact on
market conduct, which in turn causes an increase in the market price. This result
corroborates those found in Kaestner and Kahn (1990) and MacAvoy (1995). Therefore,

it appears that the market entry into the intrastate long-distance markets played the

86

opposite role that regulators had anticipated; that is, it failed to nurture the incentives of
the carriers to engage in competition.

Finally, the estimated coefﬁcient on lnACCESS is positive and signiﬁcant in both
models. In the general model lnACCESS is 0.248 and in the regulatory experiment model
it is 0.343. These estimates suggest a positive relationship between ﬂuctuations in access
fees and the market price. It should be noted that, excluding a slight increase in a few
states in 1992, access fees experienced signiﬁcant declines throughout the sample period.
The small magnitude of these estimated coefﬁcients suggest that the carriers failed to
pass through to consumers a signiﬁcant portion of the cost savings they received from

the declining price of this input.

3.6 Conclusion

The results indicate that, on average, the market conduct parameter increased as a
result of deregulation; and in turn this caused an increase in the market price. These
ﬁndings support the hypothesis that regulation restricted the market power of the long-
distance carriers.

In the ﬁrst of the two procedures this paper examines, a general model of market
conduct was estimated employing the entire panel data set. A linear approximation of
market conduct was speciﬁed as a function of regulatory, political, and structural
variables. The empirical evidence supported the ﬁmdamental premise of this paper,
which is that rate deregulation altered market conduct. It was determined that the market
conduct parameter increased 3.43 percent following rate deregulation, which in turn

caused a 4.63 percent increase in the market price.

87

 

The second procedure investigated the appropriateness of the conduct pararneter's
speciﬁcation. This required disaggregating the panel data into control and treatment
groups and examining the inter-period change in market conduct and unobserved factors.
The estimated inter-period change in the treatment group’s unobserved factors provided
evidence that the speciﬁcation of market conduct was appropriate. In particular,
unobserved factors were responsible for a very small increase in the market conduct
parameter of 0.48 percent, which caused a 0.73 percent increase in the market price.
After netting out the impact of the unobserved factors, the impact of rate deregulation on
the treatment group caused a 4.52 percent increase in the group's market conduct
parameter, which in turn caused a 7.66 percent increase in the market price.

Additional evidence that suggested non-competitive behavior was found on the
coefﬁcients for EA and IXC. In both the general and regulatory experiment models when
a state deregulated entry to the intraLATA toll market this had a positive impact on
market conduct in the interLATA market. This result suggests that state regulators
inﬂuenced market conduct in both regulated and deregulated markets. While
deregulation removed explicit constraints on price-setting conduct, implicit constraints
remained in the form of the regulator’s attitude or commitment to promoting a market
outcome. That is, entry deregulation to the intraLATA toll market signaled a
deregulatory attitude, which reduced the probability of regulatory interference and thus
enabled the long-distance carriers to more effectively exercise their market power.

The positive and signiﬁcant coefﬁcient estimates on IXC determined that markets
with a larger number of facility-based carriers produced higher prices. While this is an

unexpected (and undesirable) empirical relationship it corroborates the ﬁnding of

88

Kaestner and Kahn (1989) and MacAvoy (1995). Regardless of its interpretation it is not
consistent with competitive behavior and instead suggests that collusion may be the
underlying cause.

The model in this paper extends earlier models that examined the impact of
regulatory change on long-distance prices by explicitly considering the inﬂuence of
market conduct. While such an exercise cannot be interpreted as a complete analysis of
market power, it does uncover several results that conﬂict with what one would observe
under competition and lends empirical support to the premise that the long-distance

carriers have signiﬁcant control over the market price.

89

STATE*
1) Alabama

 

2) Arizona

3) Arkansas

4) California

5) Colorado

6) Florida

7) Georgia
8) Idaho
9) Illinois

10) Indiana

11) Iowa

12) Kansas

YEAR
1991 -94
1995-96

 

1991-96

1991-95

1 996

1991-92

1993-96

1991-96

1991-92

1993-96

1991-96

1991-96

1991-96

1991

1 992-96

1991-96

1991-92
1993-96

Appendix 3A

Regulation of Intrastate InterLATA Rates -- 1991 to 1996

RATE REGULATORY POLICY"
Price ﬂoor.

Price ﬂoor, plus mandatory pass
through of all changes in access fees.

Price cap and ﬂoor.

Flexible pricing: PSC approval of all price
changes, with a required 30 day notice
period.

Deregulated.

Rate of return.

Flexible pricing: PSC approval of all price
changes.

Price cap and ﬂoor.

Price cap and ﬂoor, plus mandatory pass
through of all changes in access fees.
Flexible pricing, with no mandatory pass
through of changes in access fees.
Regulatory policy not known by author.
Deregulated.

Price ﬂoor.

Price ﬂoor, plus mandatory pass through of
all changes in access fees.

Deregulated.

Deregulated.

Banded rate of return.

Streamlined regulation. No cost of service
studies required.

90

13) Kentucky

14) Louisiana

15) Maryland

16) Massachusetts

17) Michigan

18) Minnesota

19) Mississippi

20) Missouri

21) Montana

22) Nebraska

23) Nevada

24) New Jersey

25) New York
26) North Carolina

27) North Dakota

1991-93
1994-96

1991-94

1995-96

1 991 -93
1 994-96

1991-92
1993-96

1991
1 992-96

1 991-93
1 994-96

1 991
1 992-96

1 991 -92
1993-96

1991-94

1 995-96

1991-96

1991-94
1995-96

1 991
1 992-96

1991-96

1991-96

1991-96

Rate of return.
Streamlined regulation. Limited PSC
review of prices.

Banded rate of return.
Streamlined regulation.

Banded rate of return.
Price cap.

Rate of return.
Price cap.

Price cap.
Deregulated.

Price ﬂoor.
Deregulated.

Banded rate of return.
Price cap and ﬂoor.

Price cap and ﬂoor.
Deregulated.

Price cap, plus mandatory pass
through of all changes in access fees.
3 year deregulation "experiment."

Deregulated.

Price cap and ﬂoor.
Deregulated.

Banded rate of return.
Deregulated.

Price cap.
Price cap.

Deregulated.

91

 

28) Ohio

29) Oklahoma

30) Oregon

31) Pennsylvania

32) South Carolina
33) Tennessee

34) Texas

35) Virginia

36) Washington

37) West Virginia

38) Wisconsin

1991-93
1994-96

1991-96

1991-96

1 991 -93
1994-96

1991-96

1991-96

1991-92
1993-96

1991-96

1991-96

1991-96

1991
1992-94
1995-96

Price cap and ﬂoor.
Deregulated.

Streamlined regulation. Limited PSC
review of prices.

Deregulated.

Price cap and ﬂoor.
Deregulation.

Price cap.

Price cap.

 

Price cap and ﬂoor. I
Deregulated.

Deregulated.

Streamlined Regulation. Limited PSC
review of prices.

Regulatory policy not known by author.
Price cap.

Flexible pricing with a revenue cap.
Deregulation.

*The following 12 states are siniJe-LATA states:

 

1) Alaska

2) Connecticut

3) Delaware

4) Hawaii

5) Maine

6) New Hampshire

* It
Sources

7) New Mexico

8) Rhode Island
9) South Dakota
10) Utah

11) Vermont

12) Wyoming

The National Association of Regulatory Utility Commissioners(NARUC), Utility
Regulatory Policy in the United States and Canada, Table 100 (1991-96).

Corroboration of the NARUC report, plus additional detail on rate regulation, was
acquired from the above state's Public Service Commission staff members.

92

Appendix 3B

Non-linear Regression Results

Table 18
Non-linear Least Squares Estimates: Supply Relation

Dependent Variable: lnP,
(t-statistics in parentheses)

 

 

 

 

The General The Regulatory
Model Experiment
Control Model Treatment
Full Sample Group* Group"
(S=36, T=6) (S=11, T=4) (S=l 1, T=4)
INTERCEPT 0.529 0.945
(3.688) (1 1.928)
DEREG 0.044
(4.530)
TIME 0.004 0.102
(1.116) (2586)
EA 0.055 0.091
(1.989) (7.761)
ELECT -0.127 -0.075
(-l.652) (-2.l39)
IXC 0.111 0.156 0.067
(2.309) (5.506) (3.170)
lnACCESS 0.275 0.409
(4.505) (5.645)
STATE
DUMMY YES YES
VARIABLES
F -statistic 362.18 107.88
R-Squared 0.9238 0.9620
Adj. R-Squared 0.9213 0.9530
Instruments: lnINCOME lnINCOME
anNEMPLOY anNEMPLOY

 

 

 

 

93

 

.‘—'I.

 

Chapter Four

An Empirical Test for the Presence of Political Slack and Strategic Firm
Behavior in the Transition to Rate Deregulation in the Intrastate
Long-Distance Telecommunications Markets

The greatness of mind which manifests itself in conspicuous leadership in marshaling human and

material forces is wanted for the regulation of public utilities...Nothing is more effective in

obtaining the needed qualiﬁcations than an active and sustained public interest in regulation.
-Kerr, W.D. “Qualiﬁcations Needed for Public Utility Commissioners.” The Annals of the
American Academy, Vol.53 (1914), p.27

The typical undergraduate textbook on public utilities treats regulatory commissions a5 bodies of
automatons, selﬂessly devoting themselves to an effort to implement Smyth v. Amesl ...Economists
have come to realize that this is not an accurate description of the regulatory process. Regulators
are not automatons, but men and women who go to baseball games, advocate their political
philosophies, have their gallbladders removed, take their cats to the veterinarian, and otherwise
behave like the rest of us.

-Hilton, G. “The Basic Behavior of Regulatory Commissions.” American Economic Review,

May 1972, p.47.

4.1 Introduction

While rate deregulation today predominates in the intrastate long-distance
telecommunications markets, its achievement was gradual. During the ﬁve years
following the AT&T divestiture (1984-88) 28 of the 38 multi-LATA states2 eliminated
traditional rate-of-return (ROR) regulation, with 21 states choosing pricing-ﬂexibility
(PF) regulation (i.e., price-cap regulation, banded ROR regulation, moratorium

regulation, etc.)3 and just seven states choosing rate deregulation.4 In contrast, between

 

' 169 us. 466 (1898).

2 LATA, short for Local Access and Transport Area, is a geographic entity created at the time of divestiture
as the demarcation boundary for the Regional Bell Operating Companies (RBOCS). Calls which cross the
LATA boundary are referred to as interLATA calls and must be passed off to a long-distance carrier.
Intrastate interLATA calls are under the regulatory jurisdiction of state Public Service Commissions
(PSCs). Interstate InterLATA calls are regulated at the federal level by the Federal Communications
Commission (FCC).

3 See Kridel, Sappington, and Weisman (1996), pp. 271-274, for a detailed description of PF regulatory
policies.

94

I989-96 the pool of rate deregulated states increased from seven to 25, PF regulated
states decreased from 21 to 13, and traditional ROR regulated states decreased from 10 to
zero.5

Because the process evolved gradually, post-divestiture deregulation in the
intrastate long-distance markets provides an excellent laboratory to investigate Public
Service Commissions’ (PSC) decisions to implement regulatory change.‘5 Several recent
papers have examined the matter with regards to the initial transition from ROR to PF
regulation.7 However, no research (to the author’s knowledge) has been conducted on the
more recent transition from PF regulation to rate deregulation. Given the difference in
these two periods, it is of interest to ascertain the economic and political factors that
explain the transition to rate deregulation.

In particular, there are two distinctions between ROR and PF regulatory
environments that could produce variation in the decision to implement regulatory
change. First, the pricing ﬂexibility that the ﬁrm is endowed with under PF regulation

provides it with a strategic instrument to inﬂuence the future course of regulation. The

product of this strategic behavior is referred to as a “demonstration effect.”8 Speciﬁcally,

 

4 Between 1984-88 the following 21 states switched to a PF regulatory policy: AL, AZ, FL, GA, KS, LA,
MD, M1, MS, MO, MT, NV, NJ, NY, NC, OH, PA, SC, TN, TX, WV. The following seven states were
rate deregulated between 1984-88: ID, IL, MN, NE, OK, VA, WA.

5 The 13 PF regulated states as of year-end 1996 were AL, AZ, CO, GA, MA, MD, MS, MT, NC, NY, SC,
TN and WV.

6 With only a single exception, the decision to deregulate intrastate long-distance services has been left to
the state public service commissions. Nebraska, being the exception, was deregulated by state statute
(Legislative Bill 835) on January 1, 1987. Fortunately, for unrelated reasons, explained in section 4.4,
Nebraska is excluded from the sample.

7 Teske (1991) and Donald and Sappington (1995) examine regulatory choice in the intrastate local
telephone service markets, while Kasennan, Mayo and Pacey (1993) examine regulatory choice in the
intrastate long-distance markets.

8 For a speciﬁc discussion of the demonstration effect see Sappington and Weisman (1996). For a general
discussion of the strategic use of the political process see Ordover and Soloner (1989), pp.570-579.

95

the regulated ﬁrm’s price affects the welfare of multiple interest groups with a stake in
the regulatory process. In turn, an increase or decrease in the regulated price alters the
equilibrium political support a regulator derives from these groups (i.e., constituent votes,
campaign contributions, future employment, etc.).9 Recognizing this relationship,
namely that a regulator’s indirect utility is a function of the regulated price, the ﬁrm can
inﬂuence the regulatory process by adjusting this variable to increase the support for (or '
diminish the opposition to) its preferred outcome. Notice that ROR regulation does not

impose absolute constraints on the regulated price. Thus, in theory, a demonstration

 

effect can be observed in a ROR environment. However, the likelihood of such an
outcome is diminished by the relatively narrow pricing ﬂexibility permitted under ROR.

Second, the interim status of PF regulation contributes to the demonstration effect
in that one is more likely observe a divergence from the static (per period) proﬁt
maximizing equilibrium in an environment receptive to regulatory change. Therefore,
research on the transition from PF regulation to rate deregulation presents an issue
distinct from the previous research on the transition from ROR to PF regulation: The
regulated price and regulatory policy must be treated as endogenous variables in a model
of the regulatory process.

The objective of this paper is twofold. First, I test for the presence of a
demonstration effect in the transition from PF regulation to rate deregulation in the
intrastate long-distance markets. In doing so, I specify a simultaneous equations (Logit-
OLS) model based on a private interest theory of economic regulation. The logit

equation is based on a binary dependent variable [DEREG] that distinguishes between

 

9 See Joskow (1974) for a discussion on the inﬂuence of the regulated price on regulator behavior.

96

rate deregulated (1) and PF regulated (0) states. The equation is speciﬁed as a function of
the regulated ﬁrm’s price [PRICE]. The OLS equation examines the causal effect of
DEREG on PRICE. Therefore, a two-stage/instrumental variables procedure is
implemented to eliminate the inconsistency of the coefficients attributed to the
endogeneity of the regulated price and regulatory choice. Assuming the regulated ﬁrm
seeks rate deregulation, a necessary condition to support a demonstration effect is that the
coefﬁcient sign on DEREG in the OLS equation correspond with the coefﬁcient sign on
PRICE in the logit equation. For example, if it is determined from the logit equation that
a decline in PRICE increases the probability of rate deregulation, then to derive a
strategic interpretation a negative correlation must exist between PRICE and regulatory
policy in the OLS equation.

A consensus exists in the literature that a regulator’s private interests are integral
to predicting regulatory change.10 An open issue, however, is the degree to which
regulators operate with political slack, as suggested by Kalt and Zupan (1984).

Therefore, a second objective of the paper is to test the Kalt and Zupan theory against the
alternative hypothesis of a private interest theory of regulation without political slack, as
suggested by Stigler (1971) and Peltzrnan (1976). The test proceeds by including a
variable in the logit equation that controls for the product of slack, namely regulator
“shirking,” and it then examines the model’s response to the inclusion of this variable. A
key insight is the prediction that a decrease in slack increases the price of shirking. Thus,

as the price of shirking rises we will observe a decline in its consumption.

 

'0 See Noll (1989) for a survey of the literature.

97

 

The next section contains a brief review of the literature. Following this, in
section 4.3 I sketch a simple theoretical model that I then test. Section 4.4 describes the
estimation procedure. Section 4.5 presents the results and section 4.6 concludes. Before
proceeding I brieﬂy summarize my ﬁndings.

Two variables were included to control for the strength of an interest group’s
presence in a state. First, a positive correlation was found between the intensity of
business consumer interests, measured as the ratio of business to total local access lines,
and the probability of rate deregulation. Business customers typically favor rate
deregulation; thus, this positive relationship indicates their interests played a role in the
regulatory process. Second, the intensity of rural interests were examined by controlling
for the percentage of a state’s non-urban population. The coefﬁcient estimate on this
variable failed to achieve economic or statistical signiﬁcance. This result suggests that
rural consumers (i) do not fear their subsidized services are threatened by rate
deregulation and/or (ii) their greater presence in a state failed to produce a corresponding
increase in organized political opposition to rate deregulation.

The variable PRICE was included to test for the presence of a demonstration
effect as well as examine the inﬂuence of a group’s stake in promoting or opposing rate
deregulation. First, the negative coefﬁcient estimates on PRICE and DEREG in the logit
and OLS equations, respectively, failed to reject a demonstration effect hypothesis.
Second, the negative coefﬁcient on PRICE in the logit equation suggests that on average
a decline in PRICE produced an increase in business consumer and (possibly) regulated
ﬁrm support for rate deregulation that dominated the opposition to rate deregulation by

residential consumers.

98

 

Prior employment characteristics of the commissioners were examined by
including two variables measuring the percentage of commissioners previously employed
as (residential) consumer advocates or senior level business managers/owners. The
hypothesis is that the variables proxy for regulator bias, namely, the tendency to promote

the interests of constituent groups with which he or she has had prior afﬁliation. In turn, I

I!
argue this bias reduces the interest group’s costs in promoting support for its desired
outcome. The negative and positive coefﬁcient estimates on the respective variables
supported this hypothesis. Unfortunately, it is uncertain whether the underlying
L

 

motivation for this behavior is ideological shirking or the product of constituent
inﬂuence. Thus, the variables’ inclusion in the model was for control purposes and were
not be utilized to test for the presence of slack.

The presence of slack was conﬁrmed by a separate variable. To conduct this test I
initially included two variables to proxy for the price of shirking, deﬁned as the
percentage of elected and appointed commissioners in the ﬁnal year of their terms in
ofﬁce. The variables controlled for a speciﬁc type of shirking, namely, a regulator’s
proclivity for leisure or maintaining the status quo. Two hypothesis were tested: The
ﬁrst hypothesis asserts that political support of constituents is of greater value to the
regulator in the year he or she is running for reelection or reappointment. Consequently,
the price of shirking in this year will increase and result in a corresponding increase in the
probability of rate deregulation. This outcome is the product of three reinforcing effects:
an increase in the price of shirking generates (i) a decline in the regulator’s consumption
of status quo shirking , which causes (ii) an increase in the political support and (iii) a

decline in the political opposition to rate deregulation. The second hypothesis states that

99

variation exists in the slack of elected and appointed commissioners. The results rejected
the latter hypothesis but failed to reject the former. Thus, for efﬁciency purposes I
pooled the variables. The pooled results supporting the ﬁrst hypothesis are reﬂected in a
sizable increase in the model’s goodness of ﬁt upon inclusion of the variable: the pseudo—
Rz, increased to 0.727 from 0.601 , and a positive, statistically signiﬁcant coefﬁcient

estimate.

4.2 Literature Review

 

This section is divided into two parts: Part A reviews the theoretical literature on
economic regulation, while part B describes the empirical literature on regulatory change

as applied to the telecommunications industry.

A. The Theory of Economic Regulation

For much of this century the prevailing theory of economic regulation has been
the Public Interest (PI) theory.ll The standard assumption of PI theory is that regulators
are fully informed agents of the public interest. Given this premise, the regulatory
objective is to maximize total social surplus subject to the constraints of market demand
and supply. Clearly, a weakness in P1 (as a positive) theory derives from the assumption
that regulators are fully informed and publicly motivated. In fact, the theory has been
often criticized as a product of “normative wishings, rather than an explanation of real

1 2
world phenomena.”

 

” See Posner (1974) for a description of the evolution of P1 theory in the economics literature.
'2 See Kalt and Zupan (1984), p.279.

100

The publication of Stigler’s Theory of Economic Regulation in 1971

“crystalllze[d] a revrsronrsm 1n the economlc analys1s of regulation.” 3

The Stiglerian
theory of regulation is based on the premise that regulators are political agents, not
(necessarily) concerned with the public interest. Instead, regulation is a political process
“which contending interests seek to leverage in their pursuit of wealth.”14 Five years later
Peltzman (1976) generalized the Stiglerian theory, which in the interim had assumed the

”’5 Peltzrnan’s contribution was the

humble title “the Economic Theory of Regulation.
introduction of an explicit model of the political regulatory process whereby the
regulator’s objective function entailed maximization of “political support” (i.e., votes,
campaign contributions, future employment, etc.). From this model Peltzman derived the
result that interest groups drive the political process; success in achieving regulatory
“capture” (or inﬂuence) is predicated on a group’s inherent advantage of larger stakes and
resources, smaller size, or greater jurisdictional mobility.

Following Kalt and Zupan (1984) a reformulation of the Stigler-Peltzrnan theory
ensued.l6 The authors indicate “that approaches which conﬁne themselves to a view of
political actors as narrowly egocentric maximizers explain and predict legislative [or

”'7 The authors highlight shortcomings of the

more generally, political] outcomes poorly.
principal-agent relationship in the institution of politics: Slack in a constituent’s control

of his or her political representative - produced by policing costs, opportunism, and

 

‘3 See Peltzrnan (1976), p.211.

“ See Peltzman (1976), p.212.

'5 This title, introduced by Posner (1974), is in recognition of Stigler’s (1971) seminal contribution.

'6 The post-Kalt and Zupan (1984) literature on political “shirking” is huge. A sample of some notable
contributions include Levitt (1996), Poole and Romer (1993), Lott and Davis (1992), Bender (1991),
Levine and F orrence (1990), McCubbins, N011 and Weingast (1990), Davis and Porter (1989), Dougan and
Munger (1989), McArthur and Marks (1988), and Nelson and Silberberg (1987).

101

 

appropriability - provides the representative with an opportunity to Shirk, that is, seek
maximization of an objective function that includes non-political support variables. The
authors demonstrate that principal-agent slack plays a signiﬁcant role in explaining
political behavior and, thus, conclude that a politician’s (slack induced) shirking behavior

should be incorporated into positive models of politics.

B. Empirical Analysis of Regulatory Change: the Telecommunications Industry

Much attention has been devoted to theoretical investigations of the merits and

 

shortcomings of ROR and PF regulation.18 In addition, signiﬁcant empirical work has
examined the effects of regulatory policy on the performance of the telecommunications
industry.’9 However, there are relatively few empirical analyses of the reasons why
regulatory change is adopted in telecommunications markets.

Three recent papers - Kaserman, Mayo and Pacey (1993), Teske (1991) and
Donald and Sappington (1995) - investigate the transition from ROR to PF regulation;20
the ﬁrst examines the intrastate long-distance markets and the latter two, local telephone
service markets. They differ from the present study, however, in two fundamental
respects. First, the inﬂuence of the regulated ﬁrm’s strategic price-setting behavior, that

is the demonstration effect, is not examined. This is not surprising given the authors

 

'7 See Kalt and Zupan (1984), p.279.

'8 See, for example, Averch and Johnson (1962), Baumol and Klevorick (1970), Joskow and Schmalensee
(1986), Symposium on Price-Cap Regulation in Rand Journal of Economics (1989), Pint (1992), Liston
(1993), and Sappington (1994).

'9 See, for example, Mathios and Rogers (1989), Kaestner and Kahn (1990, 1991), Hubor, Kellogg, and
Thorne (1992), Hall (1993), Taylor and Taylor (1993), Ward (1995), MacAvoy (1995, 1996), Kahai,
Kaserman and Mayo (1996), Taylor and Zona (1997), and Knittel (1997).

102

examine regulatory transition in a ROR environment. The probability of observing a
demonstration effect under ROR regulation is diminished by the relatively strict
constraints imposed on the regulated ﬁrm’s price. Thus, a study of regulatory transition
in a PF environment offers a unique opportunity to investigate an issue that has yet to be

addressed in the literature.

Second, the empirical models used in these earlier studies are based on public

"'1 Tu. .65?

interest and private interest “political support” theories of economic regulation. In

particular, they do not consider the inﬂuence of slack on regulator conduct. This is an :

 

interesting omission given the controversy that has ensued since the publication of Kalt
and Zupan (1984). The post-Kalt and Zupan literature on political slack is substantial.
While a consensus has yet to be reached on the role of slack in the regulatory process, a
disregard for the issue deserves an explanation which the authors do not provide.

The previous studies, of course, do have an important feature in common with the
present research, namely, they all examine the same intrastate regulatory institutions (e.g.,
state Public Service Commissions). Thus, while their results may not be entirely relevant
here, they do offer guidance on constructing an empirical model and provide insight to

economic and political factors that inﬂuence the regulatory process.

 

2° Application to the electricity industry is found in Nowell and Tschirhart (1990) whom examine the
factors inﬂuencing state regulators’ compliance with the federal Public Utility Regulatory Policies Act of
1978. Oster (1980) examines some of the causes for interstate differences in social regulation.

103

4.3 A Model of Regulatory Transition
In this section I present a simple model that informs the empirical analysis that
follows. I begin by describing the model in section A, and then derive the empirical

predictions of the model in section B.

 

if
A. Description of the Model :1
The model presented in this section is based on Peltzman (1974) and Donald and
Sappington (1995). I analyze a setting with two possible regulatory regimes: PF i
E.

regulation and rate deregulation. The regulator is speciﬁed as a private interest, utility
maximizing agent, where the likelihood he or she chooses rate deregulation depends on
the level of utility under rate deregulation relative to the level of utility under PF
regulation.2' Let U (DEREG =1) denote the former and U(DEREG = 0) denote the latter.

The probability of rate deregulation is given by

P(DEREG = 1) P[U(DEREG = 1) > U(DEREG = 0)]

P[U(DEREG =1) -U(DEREG = 0) > 0] (4-1)

Because a regulator’s utility is unobserved by the econometrician, I rely on the
Stigler-Peltzman “political support” theory of economic regulation and the modiﬁcation
to this theory suggested by Kalt and Zupan as guides in specifying the variables likely to
inﬂuence the level of regulator utility associated with rate deregulation. Thus, I get the

following:

 

2' Following McFadden (1974, pp.105-142), I employ a logit model to analyze the regulator’s decision to
implement rate deregulation. See Section 4.4 for a description of the econometric model.

104

Ayn-W = U(DEREG = 1) -— U(DEREG = 0) 4 2
= n" SUP - n"OPP - U”. (R,.,,) ( i )

Equation (4.2) states that the regulator’s decision to implement rate deregulation
is predicated on (i) the support for rate deregulation exerted by group S: n5 SUP, where

n5 is the size of group S and SUP is the (net) probability that a beneﬁciary of rate

 

deregulation grants support, (ii) the opposition to rate deregulation exerted by group 0: P
n0 OPP, where n0 is the size of group 0 and OPP is the (net) probability that those
harmed by rate deregulation oppose, and (iii) the status quo utility a regulator derives
from maintaining PF regulation: U p 1:1(RSH), where RSH is the price, or opportunity cost, it

from engaging in status quo shirking activity.

Notice that a regulator’s non-political support, or “shirking,” behavior derives
from his or her unobserved personal values and motivations. Also, the impact of shirking
on the regulatory process is complex and often difﬁcult to predict. To simplify the
analysis, Up FIRSH) controls for a speciﬁc type of regulator conduct produced by slack.
The model examines shirking produced by variation in the regulators’ motivations to act
on an issue. In general, maintaining the status quo regulatory policy enables the regulator
to avoid the effort necessary in becoming informed of the alternative policy and
supervising its implementation. It may also be the case that a regulator’s inaction poses
fewer risks to his or her ﬁrture than the alternative. Therefore, the price of status quo
shirking reﬂects the net foregone political support incurred by the regulator in
maintaining PF regulation. Accordingly, an increase in RSH diminishes the utility

derived from the consumption of status quo shirking: aUpF/ORSH < 0.

105

I specify two interest groups as the principal sources of political inﬂuence: The
(group S) political support for rate deregulation is produced by business consumers and
the (group 0) political opposition to rate deregulation is generated by residential
consumers. While the producers of long-distance service clearly have a stake in the

regulatory process, the position of these parties and their relative political strength does

 

F
not vary substantially across states. As a result, I do not directly control for producer
interests. Instead, per the demonstration effect, AT&T’s is assigned an indirect role
whereby the carrier’s interests are reﬂected in its strategic manipulation of the regulated
22 3‘

price.
The probability of group S support is deﬁned as a function of x: SUP = SUP(x) ,
where x is the per capita net beneﬁt to group S from eliminating PF regulation and is

speciﬁed

IAW 3 (PRICE) — CPS 5' (rs , R,” )— C05 (”5 )
x =

(4.3)

 

S
n

The probability of group O opposition is deﬁned as a function of y: OPP = 0PP(y) ,
where y is the per capita net beneﬁt to group 0 from retaining PF regulation and is

speciﬁed

 

22 A variable measuring the number of non-regulated facility-based carriers was initially included in the
model (which I describe in section 4.4). The coefﬁcient on the variable did not achieve economic or
statistical signiﬁcance. Two shortcomings of the variable may explain this result. First, as mentioned,
there is little variation in relative strength of carrier participation across states. In fact, the two carriers that
comprise the majority share of non-regulated carriers, MCI and Sprint, experienced no change in their
market participation during the sample period. Second, the variable may not accurately control for the
carriers’ relative stakes. Clearly, a measure of market share is preferred. Unfortunately, such data is not
publicly available. Therefore, the inaccuracy of the coefﬁcient estimate may derive from a weak
correlation between a carriers’ stakes and market participation.

106

y

jaw 0 (PRICE) — CPS " (r0 , RS, )— CO" (5" )

 

U (4.4)

I assume that the per capita beneﬁts to group S from eliminating PF regulation

and the per capita beneﬁts to group 0 from retaining PF regulation are positively

correlated with the groups’ respective probabilities of support and opposition: S UP), > 0

and OPP), > 0. The variables in equations (4.3) and (4.4) are deﬁned as follows:

IAW’I - )l

CPSi( . )

COi(ni)

RSH

PRICE

= Absolute value of the Change in group i’s welfare when rate
deregulation replaces PF regulation, where ax/aIAW ”I > 0 and
ay/alow"| > 0;

= Costs incurred by group i to promote support for (or mitigate
opposition to) its preferred outcome, where ax/OCPSS < 0 and
ay/aCPS” < 0 ;

= Organizational costs incurred by group i to inform members of their
stakes and motivate activity in support of the group’s preferred
outcome, where ax/OCOS <0, ay/OCO" <0 and aCO‘/5n" >0;

= Group i speciﬁc cost shift variables;

= Price to the regulator of engaging in non-political support “shirking”
activity; and

= PF regulated price of residential long-distance telephone service.

I now brieﬂy comment on the underlying assumptions of equations (4.3) and

(4.4). First, the speciﬁcation of PRICE as a determinant of IAWS( . )| and lAWO( . )|

derives from information asymmetry and cross-subsidization of telecommunications

. 23 . .
servrces. Namely, busmess consumers are the primary source of revenues used to

subsidize other consumer groups, while residential consumers are the primary recipients

of these revenues. Consequently, the former has sought to eliminate cross-subsidization

 

23 Cross-subsidization has a long and complicated history in the telecommunications industry. Numerous
books and journal articles discuss its evolution. Notable contributions include Horwitz (1989), Temin
(I990), Brock (1994) and MacAvoy (1996).

107

by supporting the relaxation of regulatory controls and the latter, attempting to maintain
its subsidized services, has typically opposed such measures. Given the plethora of
information necessary to evaluate the true extent of the cross-subsidy, I assert that
consumers rely on a rule-of-thumb calculation to estimate the change in their welfare:
(residential) PRICE acts as an indicator of the extent of cross-subsidization. In turn, if
rate deregulation succeeds in reducing or eliminating cross-subsidies, PRICE proxies for

the extent of welfare gain to S and loss to 0 expected from rate deregulation. Thus, a

decline in PRICE suggests an increase in [AW

 

, i.e., OlAW'l/OPRICE < 0. While the

relationship is speciﬁed as having a symmetric impact on group S and 0, the magnitude
of this impact is allowed to vary across the two groups.

Notice that a demonstration effect also is dependent on information asymmetry
and is operationalized by the regulated ﬁrm’s manipulation of |AW(PRICE)|. In
particular, the regulated carrier is assumed to have complete information on its costs and,
thus, is cognizant of the true cross-subsidy between business and residential consumers.
Given this informational advantage, the regulated ﬁrm can adjust PRICE (within the
conﬁnes of the PF constraints) to exploit the trade-off between group S support and group
O opposition. For example, in an attempt to generate AUtiIity > 0, the ﬁrm could initiate
a price decrease that produced an increase in equilibrium group S support of greater
magnitude than that of group 0 opposition.

Second, the E3}. impact of group size on the beneﬁts to retaining (Cy/Ono) or

eliminating PF regulation (Ox/ans) is ex-ante indeterminate. An increase in ni provides a

108

broader base of support which necessarily reduces the per capita costs of promoting
support for group i ’s interests [CPSi( . )/n'] and potentially reduces the per capita
organization costs [C 0i(ni)/ ni]. However, an increase in ni necessarily dilutes group i’s
per capita stake in the regulatory outcome [lAWi( . )|/ n'] and increases the total
organization costs [C 0i(ni)].

Third, Yi and RSH, the determinants of CPS’( . ), measure the effect of group i’s
expenditures to promote support for their preferred outcome. As a normalization, for any
value of RSH, higher values of Yi e [0, 1] denote environments in which support
expenditures are more effective; (’BCPSi/OYi < 0. Finally, for any value of Y’, the greater
is a regulator’s disregard for group S support, the less effective is group S’s support
expenditure; since the consumption of shirking is negatively correlated with its price this
implies OCPSS/ORSH < 0. Conversely, a greater regard for the status quo suggests an
increase in the effect of group 0’s support expenditure; OCPSO/ORSH > 0.

The primary objective is to estimate the likelihood that a state chooses to adopt
rate deregulation in place of PF regulation. Substituting equations (4.3) and (4.4) into

equation (4.2) produces the following:

 

' ' " IAWS (PR'CEI ‘05" (Y5 . R55 )— C0" (5")
AUtillty = n‘ SUP "S _

(4.5)
lAw” (PR/(v51 _ CPSO (Y0 , RSI] )_ COO (n")

n” OPP[ 0 ] - U 55 (Rm )

n

 

The foregoing model predicts that rate deregulation will prevail in regulatory

settings where AUtility > 0 is more likely to hold.

109

B. Testable Implications of the Model

I now identify the factors in the model that lead to an increase or a decrease in

 

 

AUit'Iity.
S S S >
. '1' AW — CPS - C O s
Observatlon l: w = SUP — OSUP I I s, + 6C0, = 0 ; and
dn‘ 6 n‘ On‘ <
() l) ()
dA Utility aoPP IAW l— CPS ‘ C0 5C0”
— = — OPP + + = 0
dnU 6y "(I 8,10

The indeterminacy of a change in group size on regulator utility is attributed to
the uncertainty of its impact on the group’s net beneﬁts to retaining/eliminating PF
regulation. This is reﬂected in Observation 1 by the potentially conﬂicting impact of the
ﬁrst and second right-hand side components. While the sign of the ﬁrst component is
unambiguous (SUP, OPP 2 0), the sign of the second component depends on the

magnitude of the increase in support/opposition produced from a decline in CPSi/ni and

C Oi/ni , relative to the decline in support/opposition produced from a decline in IAWil/ni
and an increase in CO.

In the empirical analysis that follows, I examine the necessary condition that if
group size improves a group’s inﬂuence we will observe (i) a positive relationship
between n5 and AUtility (dAUtiIity/dnS > 0), and (ii) a negative relationship between n0
and AUtility (A Utility/an < 0).

dA Utility _ _ 55w aCPS S

dYS 6x 61’" >O’and

Observation 2:

 

dAUtility _ OOPP 5CPS" < 0

dr" 0y or"

 

110

Observation 2 holds because an increase in the effect of costs incurred to mitigate
(promote) support for rate deregulation, ceteris paribus, increases an interest group’s net
beneﬁt to maintaining (eliminating) PF regulation. The product of larger beneﬁts is

greater political opposition to (support for) rate deregulation.

dAUtility _ _ aSUP OCPSS aoPP aCPS" __ all”. >

Observation 3: +
dRSH Ox 5R SH 6y aRSH aRSH

 

 

Observation 3 states that an increase in RSH necessarily increases the effect of
support expenditures (CPSS) , while simultaneously diminishing the effect of opposition
expenditures (CPSO) and the utility derived from status quo shirking. The product of

these three responses is a necessary increase in AUtility.

V

dAUtiIity_ S 5SUP aiAWSL 0 BOPP aiAWOI _

Observation 4: _ n , n _
dPRICE 5| AW 5| 5PRICE aIAWUI OPRICE <

 

 

 

Observation 4 indicates the marginal change in group S support (lSt component)
and group O opposition (2nd component) produced by a change in PRICE. The
indeterminacy of the sign on dA Utility/dPRICE is attributed to the unknown magnitude of
the change in an interest group’s welfare produced by a change in PRICE.

Another confounding factor is the regulated ﬁrrn’s manipulation of PRICE, i.e.,
the demonstration effect. Since the 1984 divestiture, the regulated carrier AT&T has
voiced strong opposition to the rate regulation imposed on its services. Therefore, the
hypothesis is that AT&T will utilize its pricing ﬂexibility and information on the political
support group S and opposition of group O to manipulate the regulatory process. By
adjusting its price AT&T can rebalance the regulator’s equilibrium political support to

increase the probability of rate deregulation. Therefore, the effect of PRICE on AUtility,

111

be it positive or negative, may indicate the inﬂuence of the regulated carrier’s opposition
to rate deregulation.24

While dAUtility/dPRICE is ex-ante indeterminate, a negative correlation suggests
business consumer and (possibly) regulated ﬁrm support for rate deregulation, but rejects
regulatory capture by residential interest groups. However, a positive correlation
indicates the inﬂuence of residential consumer opposition and (possibly) regulated carrier

support for rate deregulation, but rejects regulatory capture by business consumers.

4.4 Empirical Analysis

The data consist of a panel of 83 observations between 1991-95 on 24 multi-
LATA states. A state was selected for inclusion if subject to PF regulation by year-end
1990. 14 of the 24 states selected for inclusion approved a rate deregulatory policy
between 1991-95.25 Note that the independent variables used to predict the probability of
rate deregulation are likely affected by the change in regulatory environment. To
eliminate this endogeneity problem I include time-series observations on the 24 PF
regulated states up to the year preceding the date rate deregulation was implemented;26
post-deregulation observations on these states are eliminated from the data set.

Therefore, the panel consists of observations on PF regulated states only; the number of

 

2" In the following section I describe the two-stage econometric techniques necessary to control for the
inconsistency attributed to the endogeneity of PRICE.

25 Of the 24 multi-LATA states included in this study 14 approved rate deregulation between 1991-95 [AR,
FL, IN, KS, LA, MI, MN, MO, NJ, NV, OH, PA, TX, and W1] and 10 retained PF regulation [A2, C0,
MA, MD, MS, MT, NC, NY, SC and TN]. Of the remaining 14 multi-LATA states excluded from the
study 10 approved rate deregulation prior to 1991 [AL, ID, IL, IA, NE, ND, OK, OR, VA, and WA], two
were not subject to PF regulation [CA, KY], and two have unknown (to the author) regulatory policies
[GA, WV].

112

 

observations on a particular state depends on when it implemented rate deregulation.27
Finally, to control for the correlation between the error terms across different years within
the same state (i.e., state ﬁxed effects) I include state dummy variables in the regression

equation.

Based on the assumption that the regulatory policy (DEREG) and the regulated

 

price (PRICE) are determined simultaneously, 1 estimate a two-equation (OLS-Logit) E:
simultaneous system. The system is speciﬁed as r
DEREO, = a, + 5, PRICE, + 5,5LINES, + a,RURAL, + a45US, + (4. 6)
5, RES, + a,APPT * LASTYR, + a7 ELECT * LASTYR, + e, 5
PRICE, = o, + b, DEREG; + b, INCOME, + b, LINES, + b, ACCESS, + u, (4.7)

where, DEREG, is a continuous variable measuring a state’s unobserved propensity to
deregulate; and the distribution of e st and us, are assumed Sech2 and standard normal,
respectively.28 Variable deﬁnitions, their descriptive statistics and data sources are
provided in Table 19.

Equation (4.6) describes the regulator’s decision making process. The selection
of the seven explanatory variables in this equation was informed by the variables in the
model of section 4.3. Equation (4.7) describes the behavior of the regulated price. I have
two comments on this equation. First, various speciﬁcations of the latter equation have

been used in empirical studies of regulated industries. From the seminal work of Stigler

 

26 If the date of implementation was after June 30, the year of implementation was included in the sample,
otherwise it was excluded.

27 For comparative purposes the model was estimated using a single cross-section of 24 observations on the
24 multi-LATA states. Similar results were found. The conclusions drawn from the model - described in
section IV - are unchanged. See appendix 4A

2" Sech2 is short for hyperbolic secant squared. See Maddala (1983), p.9.

113

 

    

Table 19
Variable Names, Deﬁnitions, and Descriptive Statistics

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Deﬁnition Mean Standard Source
Deviation
PRICE AT&T’s real rate for a daytime, 100 mile, 10 minute 2.363 0.446 a
intrastate Message Toll Service (MTS) telephone call.”
BLINES Total local business (WATS) access line divided by total 0.697 0.043 b
local access lines.+
OC C The number of facility-based long-distance carriers in a 6.790 2.637 c
state (excluding AT&T).
RURAL The percentage of a state’s population that is non-urban. 0.296 0.140 d
BUS The percentage of commissioners whose past 0.190 0.219 6
employment entailed high level business management or
ownership.
RES The percentage of commissioners whose past 0.091 0.115 e
employment entailed consumer advocacy.#
ELECT‘LASTYR The percentage of a state’s commissioners in the last 0.061 0.177 e
year of their elected term.
APPT‘LASTYR The percentage of a state‘s commissioners in the last 0.137 0.165 e
year of their appointed term.
LASTYR The percentage of a state‘s commissioners in the last 0.175 0.211 e
year of their elected or appointed term.
DEREG 1 if a state implemented rate deregulation in the 0.323 0.471 b
following period. 0 ifit retained PF regulation.
INCOME Per capita real personal income (in thousands)? 19.141 2.631
ACCESS The real access fee for a 10 minute intrastate MTS 1.177 0.382 g
telephone calla/”A
LINES The total number of local access lines divided by the 0.637 0.074 b, d
state population.

Sources: (a) Tariffed MTS rates are filed with the state Public Service Commissions;
(b) FCC, Statistics of Communications Common Carriers, Table 2.5 (1991-97);
(c) National Association of Regulatory Utility Commissioners (NARUC), Utility Regulatory Policy in the
US. and Canada (1990-96) and phone survey of commissions;
((1) US. Bureau of the Census;
(8) NARUC, Yearbook of Regulatory Agencies, Agency Proﬁles (1990-96);
(0 US. Department of Commerce, Bureau of Economic Analysis; and
(g) FCC, ARMIS REPORTS 4308 & 430].

Footnotes to Table l:

. MTS is a switched service (i.e., originates and terminates on a local telephone company’s
network), used primarily by small businesses and households. The real MTS price was calculated by
deﬂating the nominal price with the consumer price index for all goods and services (i.e., the implicit
price deﬂator).

i Wide Access Telecom. Services (WATS) is a switched service for large volume customers.

i Commissioners were categorized as consumerists if their past occupations were with the following
organizations: Ofﬁce of Attorney General (only in those states where the GAO operates as a consumer
advocacy organization), Consumer Protection, Consumer Council, Consumer Affairs, or Legal Aid
Society.

% The real price is calculated by deﬂating the nominal price with the implicit price deﬂator.

A Access fees are charged on a per minute basis and consist of four components: (1) an originating
canier common line fee; (2) a terminating carrier common line fee; (3) a trafﬁc sensitive fee; and (4) a
non-trafﬁc sensitive fee.

 

114

 

and F riedland (1962) in electricity, to Mathios and Rogers (1989) in telecommunications,
it has been standard to assume that regulatory policy is exogenous.29 Consequently, a
reduced form interpretation has been applied to equation (4.7). The novelty of the
equation’s treatment in this paper is that regulatory policy is recognized as endogenous to
the model. Thus, a two-stage estimation procedure is implemented to derive the
equation’s reduced form.

Second, notice that the endogenous variable DEREG is dependent on the ﬁrm’s
expectation of rate deregulation in the following period. Based on the assumption of

rationally formed expectations, the ﬁrm’s prediction of DEREG is the following:30

 

 

DEREG.“ = DEREG" + vs!

 

where, DEREG... is the ﬁrm’s prediction of rate deregulation in the following period,

conditional on the ﬁrm’s contemporaneous information; and E(v, )= 0 , E(v_f, )= 03 .
Therefore, I assume that on average the ﬁrm’s prediction is accurate:
E(DEREG, )= DEREG,
Given the regulated carrier’s extensive knowledge of the regulatory setting, this
assumption appears to be appropriate. Namely, in light of the voluminous amount of
public and non-public information that ﬂows between the two parties it is likely that the

regulated ﬁrm is a well informed predictor of the changes likely to occur in regulatory

poﬁcy.

 

29 This could be attributed to the strict constraints of ROR regulation, to the existing theory of economic
regulation, or as a simplifying assumption.
3" See McCafferty (1990), pp.332-335.

115

 

 

I now describe a two-stage technique for estimating a simultaneous model where
one of the dependent variables (DEREGS,) is binary.3 I The reduced form speciﬁcations

of equations (4.6) and (4.7) are the following:

DEREG; = a0 + a,BLINES_, + a, RURAL, + a,5US, + a,RES,., +
a,APPT * LASTYR, + a, ELECT * LASTrR, + a,INCOME, + (4.8)
a,LINES, + a, ACCESS, + 5,

PRICE, = 50 + 5,1NCOME, + 52 LINES, + 5, ACCESS, + 5, BLINES, .-
5,RURAL_, + 5,5US, + 5,RES, + 58APPT * LASTrR, + (4.9)
59 ELECT * LASTYR, + ,I,

 

where DEREG, is unobserved. Instead, DEREG, is observed as an indicator of DEREG, : g
DEREG, =1 if state s approved rate deregulation in period t, 0 otherwise. Therefore,
equation (4.8) becomes a logit equation. For identiﬁcation purposes, the latter three
explanatory variables in equation (4.8) are instruments for PRICE, while the latter six
explanatory variables in equation (4.9) are instruments for DEREG.
The ﬁrst stage is to estimate the reduced form logit maximum likelihood for (4.8)
and OLS for (4.9). The reduced form predicted values of PRICE, and DEREG, are
obtained and used as regressors in the structural equations (4.6) and (4.7). Finally, I

estimate (4.6) and (4.7) using logit and OLS, respectively.

A. Description of the variables: Logit Equation (4.6)
The variables BLINES and RURAL are included to control for the impact of an
interest group’s size on their political support or opposition to rate deregulation. First,

business consumers are the primary source of revenues used to subsidize other consumer

 

3' See Maddala (1983).

116

groups. They tend to pay higher rates for local service and use considerably more long-
distance during peak periods when rates are highest. Consequently, the business sector
has almost universally supported the relaxation of regulatory controls. The variable
BLINES measures the ratio of business to total local access lines in a state. To the extent

that a larger business presence in a state increases their political support for rate

 

 

deregulation, BLINES should be positively correlated with DEREG. F
Second, rural residential consumers are the major recipients of cross-subsidies in

this industry. All states currently require long-distance carriers to charge geographically 4

uniform rates. Since the cost of providing service may be greater in rural than in urban 1L

areas due to density, infrastructure and geographical considerations, such a policy
involves some cross-subsidization to the former from the latter. Thus, rural residential
customers may be opposed to rate deregulation because they fear the loss of these
subsidies. The variable RURAL measures the percentage of the state population classiﬁed
as non-urban and is included as a proxy for rural interests in the regulatory process. To
the extent that rural customers (i) fear rate deregulation will trigger the termination of
their subsidized services, and (ii) the efficacy of their political support is enhanced by
their increased presence in a state, one should observe a negative correlation between
RURAL and the probability of rate deregulation.

RES and BUS are included to control for a shift in an interest group’s support
expenditure function. They are deﬁned as the percentage of commissioners previously
employed as (residential) consumer advocates and high level managers or business
owners, respectively. The hypothesis is that prior employment controls for a regulator’s

bias to promote the interests of the constituent group with which he or she has prior

117

afﬁliation. In turn, this bias enhances the effect of each dollar spent to promote the
group’s desired outcome. I describe three factors that may explain why a regulator’s
prior employment is helpful in predicting his or her behavior. First, BUS (RES) may
control for what Gormley (1979) calls the “revolving door” hypothesis, which argues that
commissioners with business (consumer advocacy) backgrounds achieve ofﬁce with the
objective of helping their business (consumer advocacy) allies. In return, such conduct
increases the probability of securing future employment in the respective industry.
Second, the variables may proxy for the ability of business (residential) consumers to
appoint or elect commissioners that are biased in their favor. Third, the variable may
simply reﬂect the type of ideological shirking in which a regulator with a business
(consumer advocacy) aptitude chooses to engage.

Notice that, per the latter interpretation, BUS and RES potentially control for the
presence of slack. However, to conclude that slack is the underlying motivation requires
rejecting the ﬁrst two interpretations. Since each interpretation is contingent on
conditions unobserved by the econometrician, this precludes the feasibility of such a test.
Thus, the purpose for including BUS and RES is to serve as control variables and not to
corroborate a particular theory of economic regulation.

The behavior of PRICE is an obvious focal point in the regulatory process.
Changes in the variable affect the welfare of multiple interest groups, which in turn alters
the regulator’s political support. As previously speciﬁed, the coefficient on PRICE
depends on the equilibrium political support of business consumers and political

opposition of residential consumers. The regulated ﬁrm’s support is reﬂected in its

118

 

manipulation of PRICE and the effect of this manipulation on IAWS( . )l and IAW0( . )|.
A negative coefﬁcient suggests business consumer and (possibly) regulated carrier
support for rate deregulation, but rejects regulatory capture by residential interest groups.
A positive coefﬁcient indicates the inﬂuence of residential consumer opposition and
(possibly) regulated carrier support for rate deregulation, but rejects regulatory capture by
business consumers.32

Finally, to proxy for the price of regulator shirking RSI—[I include
ELEC T *LASTYR and APPT*LASTYR. Both are interaction variables that measure the
percentage of the commissioners in the last year of their term (LASTYR) multiplied by the
dummy variables ELECT and APPT, which equal one if a state elects or appoints its
commissioners, respectively. Shirking is exhibited by the variation in regulators’
motivations to act on an issue. In certain cases inaction (that is the consumption of
“leisure”) poses fewer risks to a politician’s future than the alternative. Therefore,
ELECT*LASTYR and APPT*LASTYR test for the presence of slack by controlling for the
opportunity costs a regulator incurs from the consumption of leisure shirking. The
hypothesis asserts that the utility derived from leisure consumption is time dependent.
Assuming the value a regulator places on political support increases as the date of
reelection or reappointment nears, the net gain to shirking at the beginning of a
regulator’s term in ofﬁce should exceed that in the ﬁnal year of his or her term. In short,

the consumption of leisure activity should diminish as its price increases. Distinguishing

 

’2 White (1996) discusses the inﬂuence of (residential) consumer interest groups in the recent deregulation
of intrastate electricity. Interestingly, he found that residential interests were strongly opposed to
deregulation in those states charging below average prices.

119

 

between the states that appoint and elect their commissioners does not alter the
hypothesis, however the magnitude of the coefﬁcient on ELECT *LAST YR is predicted to
exceed that on APPT *LAST YR. This latter hypothesis asserts that there is greater slack in
constituent control of appointed commissioners, relative to elected commissioners and

this, in turn, increases the price of shirking to an elected commissioner.

B. Description of the variables: OLS Equation (4.7)

Equation (4.7) examines the causal effect of different variables on PRICE, deﬁned
as the real price for an AT&T 10 minute, 100 mile, daytime, Message Toll Service (MTS)
telephone call.33

INCOME measures the state real per capita income. The hypothesis is that higher
per capita income should be associated with higher demand for long-distance service and,
thus, a higher price. This implies a positive coefficient on INCOME.

LINES measures the per capita number of local access lines in a state and is
intended to capture the efﬁciency of a state’s telecommunications infrastructure. The
hypothesis is that the size of a state’s infrastructure is positively correlated with its
efﬁciency. As the network’s efﬁciency improves, the marginal cost of providing
telephone service will decline. Therefore, the hypothesis predicts that an increase in per
capita lines will produce a decrease in PRICE.

ACCESS measures the real access fee paid by the long-distance carriers to the

local telephone company for use of the local exchange network. Access fees are an

 

’3 MTS is a switched service (i.e., it originates and terminates on a local telephone company’s network),
used primarily by residential consumers and small businesses.

120

 

integral input in the production of long-distance service. Thus, a positive coefﬁcient is
predicted on ACCESS.

DEREG is binary variable indicating states which are rate deregulated (1) and PF
regulated (0). The demonstration effect asserts that the PF regulated ﬁrm manipulates its
price to promote a deregulated outcome. This suggests the relationship between PRICE
and DEREG is contingent upon the equilibrium political support a regulator derives from
various interest groups. Consequently, the coefﬁcient on DEREG is ex-ante
indeterminate. However, for the result to be consistent with the demonstration effect it

must correspond with the coefﬁcient estimate on PRICE in the logit equation.

4.5 Empirical Results
Tables 20 and 21 report the two-stage regression results for the OLS and Logit
equations, respectively.”35 In the interest of saving space, estimated coefficients for the
state dummy variables are leﬁ out of the body of the paper and presented in appendix 4D.
To assist in the interpretation of the results reported in Table 21 note that the

probability of rate deregulation in the logit equation (4.6) is

l
P(DEREO,, = 1 1 X1.) = 6"" X": (4.10)
1 + exp X ,a

where X ,1, is the vector of explanatory variables and a is the vector of parameters:
a0 — a7 . Let 5,. denote the parameter estimate on the explanatory variable X,. Therefore,

per equation (4.10) 5,. reﬂects the extent to which an increase in the ith explanatory

 

3’ The ﬁrst-stage results for the endogenous variables PRICE and DEREG are reported in appendix 4B.

121

 

variable increases the probability that rate deregulation will be chosen relative to the

probability PF regulation is retained:

 

, 5 ﬂDEREG, = 1 I X1.)
a,=————Iog I (4.11)
5X, P(DEREG, = 0| X ,., )

Equation (4.1 1) represents the change in the probability of rate deregulation that is
produced by a marginal change in the independent variable X i , ceteris paribus. Column
2 in Table 21 reports the coefﬁcient estimates for the logit equation (4.6), while column 3
reports the £1, coefﬁcients as described in equation (4.11).

Those coefﬁcients for which a hypothesis is expressed and which attain statistical
signiﬁcance exhibit the expected sign, and the overall equations are signiﬁcant in
explaining PRICE (F-statistic = 107.61) and DEREG(ch12(30) = 62.89). Moreover, the
goodness-of-ﬁt of the equations, given by the pseudo-R2 = 0.727 in equation (4.6) and the
R2 = 0.985 in equation (4.7), is impressive.36 Finally, the percent of correct predictions in
equation (4.6) is 71.

Begin with Table 20. The coefﬁcient estimate for INCOME is 1.04 (st. error =
0.47), which conforms with the prediction that an increase in demand, produced by a rise
in per capita income, causes an increase in the price. The coefﬁcient on ACCESS is 0.45

(st. error = 0.12), which also agrees with its hypothesized effect, namely, that an increase

 

 

3’ For a related analysis of the diffusion of rate deregulation in the intrastate markets see the OLS-Tobit
results reported in appendix 4C. While similar to the regulatory choice model, the diffusion model’s focus
is on the variation between early and late adopting states.

36 If we let ln L(c’z,,I represent the value of the log-likelihood function evaluated at the parameter estimates
and In L(ci,,,o ) be the value of the log-likelihood function evaluated under the null hypothesis, i.e., the

parameters are zero, then the pseudo-R2 is given by 1— In L(&HI )/ 1n L(ci,,0 ). See StataCorp. (1995), Vol.
1, p.366.

122

Table 20
Two-Stage Estimates: OLS Equation
Dependent Variable: PRICE
(standard errors in parentheses)

 

 

 

 

Coefficient Estimates
Intercept 2.070"
(0.610)
DEREG -0.132"
(0.033)
INCOME 1.044"
(0.470)
ACCESS 0.447”
(0.120)
LINES -0.729
(0.458)
State Dummy YES
Variables
Number of Observations 83
F statistic 107.61
R-Squared 0.9849
Adjusted R-Squared 0.9757

 

 

 

 

Signiﬁcant at the 10 % level.
Signifcant at the 5 % level.

123

in the price of a key input has a positive impact on the price of long-distance service. The
coefﬁcient on LINES is —0.73 (st. error = 0.46) which is the expected sign but statistically
and economically insigniﬁcant; a 10 percent increase in the number of per capita local
access line is predicted to produce a 0.13 percent reduction in PRICE.

Finally, the coefﬁcient on DEREG is —0. 1 3 (st. error = 0.03), indicating that PSC
approval of rate deregulation produced a decline in AT&T’s price. This result, in
conjunction with the coefﬁcient on PRICE in equation (4.6), corroborates the
demonstration effect. In itself, it suggests that in the period preceding implementation of
rate deregulation, factors unrelated to supply and demand caused the state’s regulated
price to be lower than the price in (i) prior periods and/or (ii) states that retained PF
regulation.

Given their intuitive interpretation, the discussion that follows refers to the values
in column 3 of Table 21. Support for the hypothesis that group size is positively
correlated with the group’s interests is found in the positive coefficient on BLINES. In
states where the percentage of business local access lines is one standard deviation above
the sample mean, the probability of rate deregulation is 2.23 times more likely (st. error =
0.95). Since business customers are the primary contributors of cross-subsidies in this
industry, this result reﬂects business consumers strong predilection for rate deregulation.

The coefﬁcients on the interest group variables RURAL is inconclusive: it
contradicts its hypothesized interest group effect, yet fails to reach statistical signiﬁcance.
In any event, the coefﬁcient is of a fairly small magnitude; a 10 percent increase predicts

a 4 percent increase in the probability of rate deregulation. The failure of rural to

124

 

Table 21
Two-Stage Estimates: Logit Equation
Dependent Variable: DEREG
(standard errors in parentheses)+

 

 

 

 

 

The Impact of Changes in the
Explanatory Variables on the
Estimated Probability of Rate
Coefﬁcient Estimates Deregulation
Intercept 39. l 77 --
(71.336)
PRICE -6.53 1‘ -2.343"
(2.927) (1.050)
BLINES 6.225" 2.234”
(2.645) (0.949)
RURAL 1.103 0.396
(7.287) (0.2.615)
BUS 6.189" 2.221“
(2.894) (1.039)
RES 40.538” -3.781"
(4.009) (1 .439)
LASTYR 8.242 2.957
(3.017) (1.082)
State Dummies YES YES
Variables
Observations 83
Chi-Sq (30 DF) 62.89
Pseudo R-Sq 0.7266
Percent Correct 7]
Prediction

 

 

i The delta-method [See Greene p.297] used to derive st. errors in column 3.
Signiﬁcant at the 10 % level.
Signifcant at the 5 % level.

125

 

 

accurately predict rate deregulation suggests that rural customers (i) do not fear that
mandatory rate-averaging is threatened by the elimination of PF regulation and/or (ii) the
efﬁcacy of their political support is not enhanced by their increased presence in a state.

The positive coefﬁcients on BUS 2.22 (st. error = 1.04) and negative coefficient
on RES —3.78 (st. error = 1.44) indicate that the prior employment of commissioners
plays an important role in predicting regulatory change. This conﬁrms the hypothesis
that a regulator’s origins are helpful in predicting his or her current behavior. A
consequence of this predictability is that it reduces the expenditures necessary to promote
an interest group’s agenda and thus has a positive impact on their political support or
opposition to rate deregulation. Unfortunately, the underlying explanation for
this regulator bias (i.e., the revolving door, interest group inﬂuence, or regulator shirking)
is not observed. Thus, while this result is of empirical interest it cannot be used to test for
the presence of slack.37

Support for the presence of slack is found in the coefficient on LASTYR. Initially,
both ELECT*LASTYR and APPT*LASTYR were included in the logit regression.
However, the second of two hypotheses tested on the variables was rejected: there was
no evidence of greater slack in a constituent’s control of an appointed commissioner,
relative to an elected commissioner. While the coefﬁcient estimates on both variables

were positive and jointly signiﬁcant, the variation in their point estimates was statistically

 

’7 Notice that the exogeneity of BUS and RES is likely contingent upon their interpretation. To the extent
that they proxy for the inﬂuence of business and residential consumer interests the assumption may be
inappropriate. Fortunately, their inclusion in the model is not integral to achieving the paper’s objectives.
While excluding BUS and RES from the logit equation reduces its goodness-of-ﬁt (from 0.73 to 0.52), the
results are similar. Most importantly, the papers two fundamental conclusions - the presence of slack and
strategic ﬁrm behavior - are unchanged.

126

insigniﬁcant.38 This result suggests that the electoral and appointment processes were
equally (in)effective in constraining a regulator’s non-political support behavior.

For efﬁciency reasons, I pooled ELEC T*LAST YR and APPT*LASTYR,
formulating a single variable LAST YR. The restricted logit regression results including
LASTYR are reported in table 21, while the unrestricted logit regression results including
ELECT *LASTYR and APPT *LAST YR are reported in appendix 4E. LASTYR had a
marked positive impact on the model: the goodness of ﬁt, as reﬂected in the pseudo-R2,
increased to 0.727 from 0.601 upon the inclusion of the variable; and the coefﬁcients sign
was positive and statistically signiﬁcant. The latter result supports the hypothesis that a
change in the price of shirking has a negative impact on the regulator’s consumption of
shirking and the political support for rate deregulation, and a positive impact on the
political opposition to rate deregulation.

Finally, the coefﬁcient on PRICE indicates that as the variable increases above its
sample mean by one standard deviation a state became 2.34 times less likely to approve
rate deregulation (st. error = 1.05). Two conclusions are drawn from this result. First,
this outcome rejects the hypothesis of residential consumer capture, but fails to reject the
hypothesis of regulatory capture by business consumers and (possibly) the regulated ﬁrm.
Second, in conjunction with the negative coefﬁcient estimate on PRICE in equation (4.7),

this result is consistent with the demonstration effect hypothesis.

 

3" The Chi-squared statistic for the hypothesis that the coefﬁcient on APPT*LASTI’R equals the coefﬁcient
on ELECT‘LASTYR is 7.34

127

 

4.6 Conclusion

In summary, I investigated the likelihood that a state adopts rate deregulation in
place of PF regulation in the intrastate long-distance markets. An interest group rent-
seeking explanation is supported by the coefﬁcient on BLINES: the intensity of business

usage in a state, as captured in the percentage of business local access lines, is positively

 

correlated with the probability of rate deregulation. F
In addition, a test for the presence of slack is conﬁrmed. First, while the method

of achieving ofﬁce (elect vs. appoint) does not appear to produce variation in regulator

slack, the year in a regulator’s term of ofﬁce did affect his or her consumption of i

shirking. The hypothesis on the variable LASTYR asserts that the value a regulator
receives from the political support of constituents increases, and conversely the utility
derived from the consumption of shirking decreases, in the year of reelection or
reappointment. Consequently, the greater opportunity costs of maintaining the status quo
PF policy during this year increases the probability that a state chooses rate deregulation.
The positive coefﬁcients LASTYR supports this hypothesis.

I also cannot reject the demonstration effect hypothesis. The negative coefﬁcient
on PRICE in the logit equation in conjunction with the negative coefﬁcient on DEREG in
the OLS equation, suggests the strategic ﬁrm behavior plays a signiﬁcant role in
predicting regulatory change.

Finally, the prior employment characteristics of a regulator were important. The
paper examines two categories of employment: (residential) consumer advocacy (RES)
and senior level business management or ownership (BUS). The hypothesis is that a

regulator’s actions are biased towards interest groups with which he or she has had prior

128

afﬁliation. The negative coefﬁcient on RES and positive coefﬁcient on BUS support this
hypothesis. Unfortunately, it was not possible to attach an unambiguous political support
or slack interpretation to these variables. Thus, while uncovering an interesting
relationship, they do not corroborate a particular theory of economic regulation.
Although this research cannot be interpreted as a completely unambiguous test of
the theories of economic regulation, it does lend empirical support to the fundamental
premise that regulatory bodies are responsive not only to rent-seeking interest groups, but
also to their consumption demands for shirking activity. Moreover, it does help to
explain the surprisingly rapid pace of deregulation of the industry. In light of the number
of factors encouraging this trend (e. g., regulator afﬁliation, the year in their term of
ofﬁce, the intensity of business usage, and the demonstration effect), the real surprise

may be that state level long-distance telecommunications regulation is still present at all.

129

 

Appendix 4A

Estimation Results -— Cross-section Data — Regulatory Choice System

_I_)a_ta: The data comprise a single cross-section of 24 observations on the same 24 multi-
LATA states used estimate regulatory choice system [equations (4.6) and (4.7)]. Note
that the independent variables used to predict the diffusion of rate deregulation are likely
affected by the change in their regulatory environment. To eliminate this endogeneity
problem I include measures of the variables during the year rate deregulation was
approved. For those states that retained PF regulation 1 take the average of the variables

between 1991-95.

Table 22
Variable Descriptive Statistics

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Standard

Variable Mean Deviation
PRICE 2.099 0.413
BLINES 0.691 0.044
OCC 6.600 2.661
RURAL 0.288 0.139
BUS 0.199 0.229
RES 0.129 0.189
ELECT*LASTYR 0.027 0.080
APPT*LASTYR 0.180 0.192
DEREG 0.583 0.504
YEARS 1.84 1.546
INCOME 19.564 2.830
ACCESS 1 .044 0.343
LINES 0.528 0.133

 

 

 

130

 

Table 23
Two-Stage Estimates: OLS Equation
Dependent Variable: PRICE
(standard errors in parentheses)

 

 

 

Coefﬁcient Estimates

Intercept l .364
(0.933)
DEREG -0239’
(0.136)

INCOME 0.009’
(0.005)
ACCESS 0.582”
(0.194)
LINES -0.236
(0.530)

Number of Observations 24

F statistic 5.15
R-Squared 0.5754
Adjusted R-Squared 0.4636

 

 

 

 

”Signiﬁcant at the 10 % level.
Signifcant at the 5 % level.

131

Table 24
Two-Stage Estimates: Logit Equation
Dependent Variable: DEREG
(standard errors in parentheses)+

 

 

 

 

 

The Impact of Changes in the
Explanatory Variables on the
Estimated Probability of Rate
Coefﬁcient Estimates Deregulation
Intercept -1 1 .21 5 --
(12.588)
PRICE -1 .749 -0.676
(1.583) (0.612)
BLINES 19.399" 7.495”
(7.843) (3.030)
OCC -0.0714 -0.028
(0.252) (0.097)
RURAL -0.104 -0.040
(0.235) (0.091)
BUS 1.953 0.754
(2.384) (0.921)
RES 8808’ -3.403’
(4.701) (1.816)
APPT*LASTYR 10.955’ 4.232‘
(6.174) (2.385)
ELECT*LAST YR 7.169 2.770
(12.812) (4.950)
Observations 24
Chi-Sq (9 DP) 22.40
Pseudo R-Sq 0.6657
Percent Correct 64
Prediction

 

 

 

 

+ The delta-method [See Greene p.297] used to derive st. errors in column 3.
”Signiﬁcant at the 10 % level.
Signifcant at the 5 % level.

132

Appendix 48

First Stage Regression Results — Panel Data - Regulatory Choice System

Table 25
First-Stage Estimates
(standard errors in parentheses)

 

 

 

 

 

 

 

Dependent Variable = PRICE DEREG
Independent Variable Coefﬁcient Coefﬁcient
(standard error) (standard error)
Intercept 1.281 -2.991
(2.083) (7.084)
INCOME 0.020 -0.467‘
(0.022) (0.256)
ACCESS 0.516” -2.700"
(0.124) (1.173)
LINES -0898' 13.338
(0.499) (10.476)
BLINES 3.482 9.206
(3.515) (4.743)
OCC 0.031” 0.050
(0.010) (0.102)
RURAL 0.401 1.691
(1.011,). (0.430)
BUS -0.276 4.730
(0.122) (1.564).
RES 0.113 -8.510
(0.240) (3.673)
APPT*LASTYR 0.155 6.817
(0.073) (2.200)
ELECT*LASTYR 0.060 4.806
(0.089) (2.61 1)
State Dummy YES YES
Variables
Observations 83 83
F-stat / Chi-Sq 52.86 58.83
R-sq / Pseudo R-Sq 0.9796 0.7545

 

- Signiﬁcant at the 10 % level.
Signifcant at the 5 % level.

133

 

 

Appendix 4C

Estimation Results - Cross-section Data — Diffusion System

Objective: Examine the variation between those states that were early and late adopters of

rate deregulation.

Simultaneous Equations System: OLS-Tobit. The system’s speciﬁcation is identical to

 

that of the regulatory choice model except that DEREGS, - the dependent variable in (4.6)
and independent variable in (4.7) - is replaced with YEARSS, deﬁned as the years, prior to
year-end 1996, since a state implemented rate deregulation. The censored nature of
YEARSS, namely the clustering of observations around zero, requires the implementation
of a tobit estimation procedure.39 The OLS-Tobit model of diffusion is speciﬁed as

YEARS, = c, + c,PRICE, +525LINES, +c,RURAL,. +c,OCC, +

4.1C)
c, BUS, + c,CONs, + c7 ELECT * LASTYR, + c, APPT * LASTYR, + 53TERM, + n, (

PRICE, = d0 + dl YEARS, + d2 INCOME, + d3UNEMPL0Y, +
d, LINES, + dsACCESS, + w,

(4.2C)

Each equation contains one endogenous independent variable — PRICES and
YEARS 5; all others are assumed exogenous. The variables descriptive statistics are

provided in Table 22.

Data: The data comprise a single cross-section of 24 observations on the same 24 multi-

LATA states used estimate regulatory choice system [equations (4.6) and (4.7)]. Note

 

39 The tobit equation is based on a truncated distribution; that is, the dependent variable is prohibited or
censored from falling below a speciﬁed value — in this case zero. Tobin (1958) was the ﬁrst econometric
study to examine this issue. The model’s title is in recognition of the author’s contribution.

134

that the independent variables used to predict the diffusion of rate deregulation are likely
affected by the change in their regulatory environment. To eliminate this endogeneity
problem I include measures of the variables during the year rate deregulation was
approved. For those states that retained PF regulation I take the average of the variables

between 1991-95.

Estimation Technique: Maddala (1983) describes a two-stage technique for estimating a

 

simultaneous model where one of the dependent variables is censored. In condensed

notation l rewrite the system as
YEARS; = 1, PRICE, + 1,)(14 n, (4.1C)
PRICE, = 5I YEARS; + 5, X3 + w, (4.2C)
where YEARS, is a continuous variables measuring the years since deregulation; and X i-
& X f are the exogenous variables. In reduced form the system is written
YEARS, = nix, + t,, (4.3C)
PRICE, = r13 X, + at, (4.4C)
where YEARS; is censored. Instead, YEARS, is observed as an indicator of YEARS; , where

YEARS, = YEARS, if YEARS, > 0, 0 otherwise. Therefore, (4.3C) becomes a tobit equation.
The ﬁrst stage is to estimate the reduced form tobit maximum likelihood for
(4.3C) and OLS for (4.4C). The reduced form predicted values of PRICE, and YEARS;

are obtained and used as regressors in the structural models. Finally, the structural

equations are estimated using tobit for (4. l C) and OLS for (4.2C).

135

Table 26
Two-Stage Estimates: Tobit Equation
Dependent Variable: YEARS
(standard errors in parentheses)

 

 

 

 

Coefﬁcient Estimates

Intercept -7.606
(6.969)

PRICE -0275
(0.394)

BLINES 9.621 ”
(2.981)

OCC 0.410
(0.251)

RURAL -0.605
(2.794)

BUS 2.528‘
(1.273)
RES -8299"
(2.759)

ELECT*LASTYR 6.056
(7.476)

APPT*LASTYR 3.466"
(0.958)

Number of Observations 24

Chi2(9 DF) 28.67
Pseudo R-Squared 0.3261

 

 

 

 

‘ Signiﬁcant at the 10 % level.
Signifcant at the 5 % level.

136

Table 27
Two-Stage Estimates: OLS Equation
Dependent Variable: PRICE
(standard errors in parentheses)

 

 

 

Coefﬁcient Estimates

Intercept l .736
(1.049)
YEARS -0.060
(0.052)
INCOME 1.017"
(0.581)
ACCESS 0.533”
(0.209)
LINES -0.109
(0.545)

Number of Observations 24

F statistic 4.42
R-Squared 0.5379
Adjusted R-Squared 0.3931

 

 

 

 

 

Signiﬁcant at the 10 % level.
Signifcant at the 5 % level.

137

Estimation Results — Panel Data - Regulatory Choice System State Fixed Effects

Appendix 4D

Table 28

Two-Stage Estimates: State Dummy Variables+

(standard errors in parentheses)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Logit Equation (3) OLS Equation (4)
1. Arizona Omitted Omitted
2. Arkansas 0.901 0.196"
(1.689) (0.068)
3. Colorado -5.132 -0.105
(3.120) (0.095)
4. Florida -1.150 -0.169
(6.232) (0.191)
5. Indiana 0.683 -0.473'
(2.661) (0.255)
6. Kansas -8.701 -0.009
(5.973) (0.162)
7. Louisiana 86.499" -0.357"
(15.384) (0.078)
8. Massachusetts -9.559 0.280
(11.107) (0.357)
9. Maryland -29.l86w -0.091
(12.720) (0.312)
10. Michigan 0.741 0.662"
(1.556) (0.171)
11. Minnesota -2.112 0.112
(2.040) (0.233)
12. Mississippi 1.653 0497"
(7.822) (0.122)
13. Missouri 0.980 0.282"
(3.213) (0.133)
14. Montana 6.878' -0.286"
(3.767) (0.066)
15. New Jersey -25.401‘ 0.169
(13.790) (0.435)
16. New York -4.987 -0.168'
(6.007) (0.088)
17. Nevada -29.076" -0.084
(11.154) (0.238)
18. North Carolina 0673 -0.097
(0.359) (0.108)
19. Ohio 1.275 0.154
(4005) (0.142)
20. Pennsylvania 1.232 0.551"
(5.765) (0.207)
21. South Carolina 2.914 0.019
(4.100) (0.106)
22. Tennessee 1.001 0.218
(0.759) (0.665)
23. Texas -9.870 -0.043
(4.114) (0.126)
24. Wisconsin 2.863 -0.012
(4.980) (0.145)

 

 

 

+ While the F statistic of 43. 12 conﬁrmed the hypothesis that state ﬁxed effects are jointly signiﬁcant in equation (4),

the Chi-squared statistic of 7.41 for equation (3) fails to reject the joint signiﬁcance of state ﬁxed effects.
" Signiﬁcant at the 10 % level; " Signifcant at the 5 % level.

138

 

 

Appendix 4E

Estimation Results — Panel Data - Unrestricted Logit Equation

Table 29

Two-Stage Estimates: Unrestricted Logit Equation
Dependent Variable: DEREG

(standard errors in parentheses)+

 

The Impact of Changes in the
Explanatory Variables on the
Estimated Probability of Rate

 

 

 

 

Coefﬁcient Estimates Deregulation
Intercept 19.673 --
(39.099)
PRICE -3654" -1.131**
(1.714) (0.531)
BLINES 6.012“ 2.157"
(2.61 1) (0.937)
RURAL -0.707 -0.254
(4.719) (1.693)
BUS 6.310" 2.221 "
(2.733) (1.039)
RES 45.644" -3781“
(8.409) (1.439)
ELECT*LASTYR 7.928" 2.845“
(4.082) (1.465)
APPT*LASTYR 6.503” 2.333"
(2.414) (0.866)
State Dummies YES YES
Variables
Observations 83
Chi-Sq (31 DF) 62.74
Pseudo R-Sq 0.7249
Percent Correct 69

Prediction

 

 

+ The delta-method [See Greene p.297] used to derive st. errors in column 3.
* Signiﬁcant at the 10 % level.
" Signifcant at the 5 % level.

139

 

 

REFERENCES

Averch, H. and Johnson, L. “Behavior of the Firm Under Regulatory Constraint.”
American Economic Review, Vol.52, No.5 (1962), pp.1052-1069.

Baumol, W. and Klevorick, A. “Input Choices and Rate-of-Retum Regulation: An
Overview of the Discussion.” Bell Journal of Economics, Vol.1, No.2 (1970), pp.162-
190.

Becker, G. “A Theory of Competition among Pressure Groups for Political Inﬂuence.”
Quarterly Journal of Economics, Vol.98 (1983), pp.37l-400.

Bender, Bruce A. “The Inﬂuence of Ideology on Congressional Voting.” Economic
Inquiry, Vol.29, No.3 (1991), pp.416-428.

Bresnahan, T.F. "Empirical Studies of Industries with Market Power." In Handbook of
Industrial Organization. Edited by R. Schmalensee and R. Willig. North Holland
(1989), pp.1011-1057.

Brock, G.W. Telecommunications Policy for the Information Age: From Monopoly to
Competition, Harvard University Press (1994).

Card, D., and Krueger, A.B., Myth and Measurement: The New Economics of the
Minimum Wage. Princeton, NJ: Princeton University Press (1995).

Davis, M. and Porter, P.K. “A Test for Pure or Apparent Ideology in Congressional
Voting.” Public Choice, Vol.60, No.2 (1989), pp.101-112.

Donald, S.G., and Sappington, D.E.M. “Explaining the Choice Among Regulatory Plans
in the US. Telecommunications Industry.” Journal of Economics and Management
Strategy, Vol.4, No.2 (1995), pp.237-265.

Dougan, W.R. and Munger, MC. “The Rationality of Ideology.” Journal of Law and
Economics, Vol.32 (1989), pp] 19-142.

Edelman, SA. “The FCC and the Decline in AT&T’s Long Distance Residential Rates,
1980-1992: Did Price Caps Do It?” Review of Industrial Organization, Vol.12 (1997),
pp.537-553.

Federal Communications Commission, Statistics of Communications Common Carriers.
Industry Analysis Division, Washington, DC. (1991-1997).

, Reference Book: Rates, Indexes, and Household Expenditure for Telephone
Services. Industry Analysis Division, Washington, DC. (1997a).

140

, Trends in Telephone Service. Industry Analysis Division, Washington, DC.
(1997b)

Flarnm, K. “Technological Advance and Costs: Computer versus Communications.” In
Changing the Rules: Technological Change, International Competition, and Regulation
in Communications. Edited by R.W. Crandall and K. Flamm. The Brookings Institution
(1989), pp.13-61.

Garﬁnkel, L. “The Growth of Competition in the US. Long-Distance Industry.”
Telecommunications Policy, July (1993), pp.323-330.

Gatto, J .P., Kelejian, H.H., and Stephan, S.W. "Stochastic Generalization of Demand
Systems with an Application to Telecommunications." Information Economics and

Policy, Vol.3 (1988), pp.283-309.

Gatto, J .P., Langin-Hooper, J ., Robinson, RB, and Tyan, H. "Interstate Switched Access
Demand Analysis." Information Economics and Policy, Vol.3 (1988), pp.333-358.

Gelfand, MD. and Spiller, P.T. “Entry Barriers and Multiproduct Oligopolies.”
International Journal of Industrial Organization, Vol.5 (1987), pp.101-113.

Genesove, D. and Mullin, W.P. “Testing Static Oligopoly Models: Conduct and Cost in
the Sugar Industry, 1890-1914.” Rand Journal of Economics, forthcoming (1998).

Geroski, P.A. “In Pursuit of Monopoly Power: Recent Quantitative Work in Industrial
Economics.” Journal of Applied Econometrics, Vol.3 (1988), pp.107-123.

Gormley, W.T., Jr. “A Test of the Revolving Door Hypothesis at the FCC.” American
Journal of Political Science, Vol.23, No.4 (1979), pp.665-683.

Greene, W. Econometric Analysis (2nd Edition). New York: MacMillan Publishing
Company (1993).

Gruber, J. "The Incidence of Mandated Maternity Beneﬁts." The American Economic
Review, Vol.84, No.3 (1994), pp.622-641.

Hall, R.E. Long Distance: Public Beneﬁtsﬁ'om Increased Competition. Applied
Economics Partners, Menlo Park, CA (1993).

Harris, RC. and Kraﬁ, C.J. “Meddling Through: Regulating Local Telephone
Competition in the United States.” Journal of Economic Perspectives, Winter (1997).

141

 

Hausman, J .A. “Competition in Long-Distance and Telecommunications Equipment
Markets: Effects of the MF J” Managerial and Decision Economics, Vol.16 (1995),
pp.365-3 83.

Hausman, J .A., Tardiff, T. and Belinfante, A. “The Breaking up of AT&T and Changes in
Telecommunications Regulation: What Are The Lessons?” American Economic Review,
Vol.83, No.2 (1993), pp. 178-84.

Horwitz, RE. The Irony of Regulatory Reform: the deregulation of American
Telecommunications. New York: Oxford University Press (1989).

Huber, P.W. “Telephones, Competition, and the Candice-Coated Monopoly.” Regulation,
No.2 (1993a), pp.34-43.

Huber, P.W. “AT&T Leads, Competition Follow.” Cato Review of Business and
Government Regulation, No.3 (1993b), pp.4-6.

Huber, P.W., Kellogg, M.K., and Thorne, J. The Geodesic Network II: 1993 Report on
Competition in the Telephone Industry (1992).

Jacquemin, A. and Slade, M.E. “Cartels, Collusion, and Horizontal Merger.” In
Handbook of Industrial Organization. Edited by R. Schmalensee and R. Willig. North
Holland (1989), pp.415-473.

Joskow, P. “Inﬂation and Environmental Concern: Structural Change in the Process of
Public Utility Price Regulation.” Journal of Law and Economics, Vol.17 (1974), pp.291-
327.

Joskow, P. and Schmalensee, R. “Incentive Regulation for Electric Utilities.” Yale
Journal on Regulation, Vol.4, No.1 (1986), pp.1-49.

Kaestner, R. and Kahn, B. "The Effects of Regulation and Competition on the Price of
AT&T Intrastate Telephone Service." Journal of Regulatory Economics, Vol.2 (1990),
pp. 1-1 5.

Kaestner, R. and Kahn, B. “The Impact of IntraLATA Competition on Local Exchange
Company Prices” in Economic Innovations in Public Utility Regulation, Edited by Crew
(1991)

Kahai, S.K., Kaserman, D.L., and Mayo, J .W. “Deregulation and Predation in Long-

Distance Telecommunications: An Empirical Test.” The Antitrust Bulletin, Fall (1995),
pp.645-666.

142

Kahai, S.K., Kaserman, D.L., and Mayo, J .W. “15 the ‘Dominant Firm’ Dominant? An
Empirical Analysis of AT&T’s Market Power.” Journal of Law and Economics,
Vol.XXXIX, October (1996), pp.499-517.

Kahn, AB and Shew, W.B. “Current Issues in Telecommunications Regulation:
Pricing.” Yale Journal on Regulation (1987).

Kalt, J. and Zupan, M. “Capture and Ideology in the Economic Theory of Politics.”
American Economic Review, Vol.74 (1984), pp.279-300.

Kalt, J. and Zupan, M. “The Apparent Ideological Behavior of Legislators: Testing
Principal-Agent Slack in Political Institutions.” Journal of Law and Economics, Vol.33,
No.1(1990), pp.103-l31.

Kaserman, D.L. and Mayo, J .W. “Long-Distance Telecommunications: Expectations and
Realizations in the Post-Divestiture Period.” In Incentive Regulation for Public Utilities,
edited by MA. Crew. Boston, MA: Kluwer Academic Publishers (1994), pp.83-l 1 l.

Kaserman, D.L., Mayo, J .W., and Pacey, P.L. “The Political Economy of Deregulation:
The Case of Intrastate Long Distance.” Journal of Regulatory Economics, Vol.5 (1993),
pp.49-63.

Keller, J .J . “US West Strikes Marketing Alliance with Qwest in Bold Move Skirting
Rules.” Wall Street Journal, May 7, 1998, p.A2, A8.

Knittel, C.R. “Interstate Long Distance Rates: Search Costs, Switching Costs, and Market
Power.” Review of Industrial Organization, Vol.12 (1997), pp.519-536.

Kraushaar, J .M. Fiber Optic Update. Federal Communications Commission, Common
Carrier Bureau, Industry Analysis Division, Washington, DC. (1997).

Kridel, D.J., Sappington, D.E.M., and Weisman, D.L. “The Effects of Incentive
Regulation in the Telecommunications Industry: A Survey.” Journal of Regulatory
Economics, Vol.9 (1996), pp.269-306.

Kwoka, J .E. "Implementing Price Caps in Telecommunications." Journal of Policy
Analysis and Management, Vol.12, No.4 (1993), pp.726-752.

Levine, ME. and Forrence, J .L. “Regulatory Capture, Public Interest, and the Public
Agenda: Towards a Synthesis.” Journal of Law Economics and Organization, Vol.6
(1990), pp.167-l98.

Levitt, How Do Senators Vote? Disentangling the Role of Voter Preferences, Party

Afﬁliation, and Senator Ideology.” American Economic Review, Vol.86, No.3 (1996),
pp.425-441.

143

Liston, C. “Price-Cap Versus Rate-of-Retum Regulation.” Journal of Regulatory
Economics, Vol.5, No.1 (1993), pp.25-48.

Lott, J .R. and Davis, M.L. “A Critical Review and Extension of the Political Shirking
Literature.” Public Choice, Vol.74, No.4 (1992), pp.133-150.

Loube, R. and Pilalis, L.E. "State Experience in InterLata Toll Regulation." Journal of
Economic Issues, Vol.28, No.2 (1994), pp.415-425.

MacAvoy, P.W. The Failure of Antitrust and Regulation to Establish Competition in
Long-Distance Telephone Service. Cambridge, MA: The MIT Press and Washington,
DC: The AEI Press (1996).

MacAvoy, P.W. “Tacit Collusion Under Regulation in the Pricing of Interstate Long-
Distance Telephone Services.” Journal ofEconomics and Management Strategy, Vol.4,
No.2 (1995), pp.147-185.

Mackinnon, J .G., White, H., and Davidson, R. "Tests For Model Speciﬁcation in the
Presence of Alternative Hypotheses." Journal of Econometrics, Vol.2] (1985), pp.53-70.

Maddala, G.S., Limited Dependent and Qualitative Variables in Econometrics,
Cambridge University Press, Cambridge (1983).

Madrian, B. "Employment-Based Health Insurance and Job Mobility: Is There Evidence
of Job-Lock?" The Quarterly Journal ofEconomics, February (1994), pp.27-54.

Mathios, AD. and Rogers, PR. "The Impact of Alternative Forms of State Regulation of
AT&T on Direct-Dial, Long-Distance Telephone Rates." Rand Journal of Economics,
Vol.20, No.3 (1989), pp.53-68.

McArthur, J .R. and Marks, S.V. “Constituent Interests Versus Legislator Ideology: the
Role of Political Opportunity Cost.” Economic Inquiry, Vol.26, No.3 (1988), pp.461-
470.

McCafferty, S. Macroeconomic Theory. New York, NY: Harper & Row, Publishers, Inc.
(1990)

McChesney, F .8. “Rent Extraction and Rent Creation in the Economic Theory of
Regulation.” Journal of Legal Studies, Vol.16 (1987), pp.101-118.

McCubbins, M.D., Noll, R., and Weingast, B. “Structure and Process, Politics and Policy:

Administrative Arrangements and the Political Control of Agencies.” Virginia Law
Review, Vol.75 (1989), pp.431-82.

144

McCubbins, M.D., Noll, R., and Weingast, B. “Slack, Public Interest, and Structure-
Induced Policy.” Journal of Law Economics and Organization, Vol.6 (1990), pp.203-
212.

McFadden, D. “Conditional Logit Analysis of Qualitative Choice Behavior.” In
Frontiers in Econometrics. Edited by P. Zarembka. New York: Academic Press (1974).

Nelson, D. and Silberberg, E. “Ideology and Legislator Shirking.” Economic Inquiry,
Vol.25, No.1 (1987), pp.15-25.

Noam, E.M. "Assessing the Impacts of Divestiture and Deregulation in
Telecommunications." Southern Economic Journal, Vol.59 (1992), pp.438-449.

Noll, A.M. “The Costs of Competition: FCC Telecommunications Orders of 1997.”
Telecommunications Policy, Vol.22, No.1 (1998), pp.47-56.

Noll, R.G. “Economic Perspectives on the Politics of Regulation.” in Handbook of
Industrial Organization, R. Schmalensee & R.D. Willig, eds., Vol.2, Ch.22 (1989)
pp.1253-1287.

Noll, R.G. “State Regulatory Responses to Competition and Divestiture in the

Telecommunications Industry.” In Antitrust and Regulation, edited by Grieson. Boston:
Kluwer Academic Publishers (1986), pp.165-200.

Nowell, C. and Tschirhart, J. “The Public Utility Regulatory Policy Act and Regulatory
Behavior.” Journal of Regulatory Economics, Vol.2, No.1 (1990), pp.21-36.

Ordover, J .A. and Soloner, G. “Predation, Monopolization, and Antitrust.” in Handbook
of Industrial Organization, R. Schmalensee & R.D. Willig, eds., Vol.1, Ch.9 (1989) pp.
537-596.

Oster, S.M. “An Analysis of Some Causes of Interstate Differences in Consumer
Regulations.” Economic Inquiry, Vol.18, No.1 (1980), pp.39-54.

Peltzrnan, S. “Towards a More General Theory of Regulation.” The Journal of Law and
Economics, Vol.19 (1976), pp.211-240.

Pint, E.M. “Price-Cap Versus Rate-of-Retum Regulation in a Stochastic-Cost Model.”
Rand Journal of Economics, Vol.23, No.4 (1992), pp.564-578.

Poole, K.T. and Romer, T. “Ideology, ‘Shirking,’ and Representation.” Public Choice,
Vol.77, No.1 (1993), pp.185-196.

Porter, Michael “Competition in the Long Distance Telecommunications Market.”
Boston, MA: Monitor Company (1993).

145

 

Posner, R.A. “Theories of Economic Regulation.” Bell Journal of Economics, Vol.5
(1974), pp.335-358.

Rubinovitz, R.N. “Market Power and Price Increases for Basic Cable Service Since
Deregulation.” Rand Journal of Economics, Vol.24, No.1 (1993), pp.1-18.

Sappington, D.E.M. “Designing Incentive Regulation.” Review of Industrial
Organization, Vol.9, No.3 (1994), pp.245-272.

Sappington, D.E.M. and Weisman, D. “Potential Pitfalls in Empirical Investigations of
the Effects of Incentive Regulation.” Information Economics and Policy (1996).

Scherer, F .M. and Ross, D. Industrial Market Structure and Economic Performance.
Boston, Mass: Houghton Mifflin Company (1990).

Schoenwald, S.M. "Regulating Competition in the Interexchange Telecommunications
Market." Federal Communications Law Journal, Vol.49, No.2 (1997).

Sievers, M. “The Law and Economics of Intralata Competition: 1+ Issues and Access
Charge Imputation.” In Competition and the Regulation of Utilities, edited by MA.
Crew. Boston: Kluwer Academic Publishers (1991), pp.7-31.

Spiller, P.T. and Favaro, E. “The Effects of Entry Regulation on Oligopolistic
Interaction: The Uruguayan Banking Sector.” Rand Journal of Economics, Vol.15, No.2
(1984), pp.244-254.

StataCorp. Stata Statistical Software: Release 4.0, College Station, TX: Stata
Corporation.

Stigler, G.J. “The Theory of Economic Regulation.” Bell Journal of Economics and
Management Science, Vol.2 (1971), pp.3-21.

Stigler, G.J. and Friedland, C. “What Can Regulators Regulate? The Case of Electricity.”
Journal of Law and Economics, Vol.5 (1962), pp.1-l6.

Symposium on Price-Cap Regulation, Rand Journal of Economics, Vol.20, No.3 (1989),
pp.369-472.

Taylor, L.D. Telecommunications Demand in Theory and Practice, Kluwer Academic
Publishers (1994).

Taylor, W.E. and Taylor, L.D. “Postdivestiture Long-Distance Competition in the United
States.” American Economic Review, Vol.83 (1993), pp.185-190.

146

Taylor, W.E., and Zona, J .D. “An Analysis of the State of Competition in Long-Distance
Telephone Markets.” Journal of Regulatory Economics, Vol.1] (1997), pp.227-255.

Temin, P. “Cross Subsidies in the Telephone Network aﬁer Divestiture.” Journal of
Regulatory Economics, Vol.2 (1990), pp.349-362.

Teske, P.E. “Rent-Seeking in the Deregulatory Environment: State
Telecommunications.” Public Choice, Vol.68 (1991), pp.235-243.

Ward, M.R. “Measurements of Market Power in Long Distance Telecommunications”
Staff Report. Washington, DC, Federal Trade Commission, Bureau of Economics
(1995)

Ward, M.R. "Prices, Productivity and Industry Structure: Evidence From the Long-
Distance Industry." University of Illinois Dept. of A griculture and Consumer Economics
working paper, May (1996).

Waverman, L. “US. Interexchange Competition.” In Changing the Rules:
Technological Change, International Competition, and Regulation in Communications.

Edited by R.W. Crandall and K. Flamm. The Brookings Institution (1989), pp.62-113.

White, M. “Power Struggles: Explaining Deregulatory Reforms in Electricity Markets.”
Brookings Papers on Economic Activity: Microeconomics (1996), pp.201-250.

147